…ultisig#4412)

* refactor(send): split submit functions into per-defiType strategies

Phase 3 of the SendFormViewModel refactor (PR 1 vultisig#4377, PR 2 vultisig#4393 already
merged). Each defiType branch in SendFormViewModel now lives behind a
dedicated SendSubmitStrategy implementation, so the VM no longer carries
the 8 submit functions, the 5 create*DepositTransaction helpers, the
inline accountValidation/awaitGasFee/getFeesFiatValue/getBitcoinTransactionPlan
helpers, or the AccountValidation data class. SendFormViewModel.kt drops
from 3261 → 1879 LOC.

The strategies are plain classes manually constructed in the VM init block
(matching the AddressManager / AmountManager precedent from PR 2). No Hilt
multibinding — defiType is a closed enum and one-class-per-concern keeps
the diff lift-and-shift. Public surface preserved: send() stays as a
one-line passthrough so the existing screen bindings and characterization
tests keep working; the four previously-private submit functions
(stake/mint/redeem/withDrawUSDCCircle) and the unused public bond/unbond/
unstake passthroughs are deleted because nothing outside the VM ever
called them. AccountValidator and BitcoinPlanService come out as small
shared collaborators (the latter still has one VM-side caller in the
collectPlanFee flow).

Adds 29 Phase C tests covering AccountValidator's validate() error
branches plus one error-path test per strategy (no_address, no_amount,
RUNE-balance-insufficient, operator-fee out of range, slippage,
WITHDRAW_RUJI silent early-return, DefaultSend pre-launch
expandSection+focus). Total send.* test count: 47 Phase A → 47, Phase B
managers 25 → 25, Phase C strategies 0 → 29 = 135 tests. All green via
./gradlew :app:testDebugUnitTest --tests "com.vultisig.wallet.ui.models.send.*".

Also clears the IDE warnings the linter surfaced on the trimmed VM:
unused BondAddress enum value, unused calculateGasLimit/enableAdvanceGasUi
helpers, the cascade if-else in loadAccounts (now a when over defiType),
the redundant R.string qualifier on one error-path message, and the
missing @param: use-site target on chainLogo.

Percentage/calculate functions (chooseMaxTokenAmount, choosePercentageAmount,
calculatePercentageWithAccurateFee, fetchPercentageOfAvailableBalance) are
deliberately deferred — they share the defiType switch with submit but
also write to gasFee/specific and orchestrate fee recalculation, a
different concern that would double the size of this PR.

Co-Authored-By: aminsato <Amin.saradar@yahoo.com>

* chore(send): drop unused bond/unbond/unstake VM passthroughs

These three public methods previously existed only so the in-VM
onClickContinue() switch could call them by name. After the strategy
extraction, onClickContinue dispatches directly to the strategies and
nothing outside the VM ever called bond/unbond/unstake — confirmed by
grep across app/src/. Keeping send() because the existing
SendFormViewModelSubmitEvmTest characterization tests still call it.

Co-Authored-By: aminsato <Amin.saradar@yahoo.com>

* fix(send): address CodeRabbit review on Phase 3 strategies

Five fixes raised in PR vultisig#4412 review:

1. Rethrow CancellationException in every strategy submit catch block (and in
   the inner resolveName/resolveDstAddress catches in BondStrategy,
   UnbondStrategy, DefaultSendStrategy, AccountValidator). The broad
   catch (e: Exception) was converting normal coroutine cancellation into a
   spurious error toast instead of letting it propagate.

2. DefaultSendStrategy: drop the redundant first chainAccountAddressRepository
   .isValid check on the raw input. The check on the resolved dstAddress a few
   lines later already covers the validation, and the raw-input check broke
   the corner case where the user types an ENS/thorname and taps send before
   AddressManager's debounce has run — the first check would reject the name
   before resolveDstAddress could translate it.

3. UnstakeStrategy WITHDRAW_RUJI branch: replace the silent `?: return@launch`
   when the RUJI account lookup misses with `throw InvalidTransactionDataException(
   send_error_no_token)`. Previously tapping Continue with no RUJI account just
   appeared to do nothing — now the user sees an error.

4. UnstakeStrategy: wrap the getSpecific calls in createRUJIRewardsDeposit
   Transaction and createYTCUnstakeDepositTransaction with `withContext(
   Dispatchers.IO)`. Only createRUJIUnstakeDepositTransaction had it
   previously; the other two were running blocking repository calls on the
   caller dispatcher.

5. UnstakeStrategy: replace the BigInteger→Double basis-points calculation in
   createYTCUnstakeDepositTransaction with integer math
   (tokenAmountInt * 10_000 / totalTokenAmount) so token-sized values don't
   lose precision and emit the wrong unstake memo.

Updates UnstakeStrategyTest's WITHDRAW_RUJI test from "returns silently when
no RUJI account exists" to "surfaces no_token error" to match vultisig#3. All 135
send.* tests still green.

Co-Authored-By: aminsato <Amin.saradar@yahoo.com>

* refactor(send): address review comments on submit strategies

Code:
- enumerate every DeFiNavActions in onClickContinue so a future variant
  fails to compile instead of silently routing through DefaultSendStrategy
- drop the test-only fun send() passthrough; tests use onClickContinue()
- DefaultSendStrategy now goes through accountValidator.validate() and
  wraps getSpecific in Dispatchers.IO so the regular send path matches
  every other strategy
- guard every strategy submit() with a Job? so a fast double-tap cannot
  launch two concurrent submits and persist a duplicate transaction
- replace error()/RuntimeException paths in Mint/Redeem/Stake/Unstake/
  WithdrawUsdcCircle with InvalidTransactionDataException so the dialog
  no longer leaks raw English strings or enum identifiers
- bind nonDeFiAccount to a non-null val in WithdrawUsdcCircleStrategy
  so the later .token access is safe under any future guard tweak
- swap the UiText.Empty fallback in every catch (e: Exception) for
  R.string.dialog_default_error_body so the dialog body is never blank

Tests:
- add a positive happy-path test to every strategy that reaches
  addTransaction(...) and asserts on the produced Transaction or
  DepositTransaction body — memo, dstAddress, contract address, and
  the ruji/tcy/ytcy/yrune contract selection inside the variant flows
- de-relax the accountsRepository mock in WithdrawUsdcCircleStrategyTest
  and stub loadAddresses per test so future fixtures actually exercise
  the nonDeFiBalance < gasFee guard
- mockkObject(ThorchainFunctions) and mockkObject(EthereumFunction) so
  the JNI-bound encoders don't require Trust Wallet Core's native
  library at JVM unit-test time

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: hailhydra <johnnyluo1980@gmail.com>
Co-authored-by: Roman <32820910+rkokhatskyi@users.noreply.github.com>
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>