fix(security): bounds-check length prefixes on decode

Untrusted flatty input (bytes off the network) could name a length far
larger than the buffer. The seq/string/table/set decoders trusted it and
setLen/copyMem'd against it -- a hostile packet triggered either a huge
allocation (OOM) or a copyMem heap over-read (SIGSEGV, which no try/except
catches). Validate every count against the bytes that actually remain
before allocating or copying; reject bad buffers with a new FlattyError
(a CatchableError) so trust boundaries drop the input instead of crashing.
Holds even under -d:danger, where Nim's own bounds checks are off.

Bump to 0.4.1.

Bump version to 0.4.0

Merge pull request treeform#30 from guzba/master

readStr + fix

v0.3.0 - Add flatty/encode

Merge pull request treeform#28 from guzba/master

0.2.4 ptr UncheckedArray[uint8] binny reading

Merge pull request treeform#26 from guzba/master

updated github workflow

Merge pull request treeform#23 from guzba/master

fix for nim issue #10128