v0.3.0-jinx7

Fix the crate publishing workflow

v0.3.0-jinx6:

- Fix release workflow

v0.3.0-jinx5 release

- Fixes docker image

v0.3.0-jinx4 release

This is quite a small release as the code inclusion bug was really nasty...
and prevented any possible development using `jinko`. A bunch of other bugs
regarding generics have been fixed.

- #474

Multiple instructions such as `VarAssign` and `FieldAssign` did not expand
generics properly. This has been fixed in the PR aiming for generic method
calls to resolve correctly.

- #475

Previously, non-generic types were getting generically-mapped during the
expansion pass. This meant that code like this:
```rust
type Something;
func takes_generic_and_some[T](one: T, two: Something) {}
```
would error out for not mapping the generic type `Something` properly.

- #477

Typechecking had broken test functions by forbidding the user to declare
a test with the same name as an existing function. This is now fixed. Test
functions are also now typechecked properly, which wasn't the case before.

- #478

You are no longer allowed to perform generic function calls if the function
does not contain generics.

```rust
func no_gen(a: int, b: string) -> float {
    15.4
}

no_gen[int, string, bool](15, true); // Error
```

- #480

`fmt_int` has been reimplemented completely in jinko! There is a small
performance hit, and the implementation is extremely ugly due to a weird
recursion argument mapping bug and the lack of switch cases/pattern matching.

The implementation was benchmarked against a similarly ugly python
implementation, and a call to the previous version of `fmt_int()`, which is
simply a wrapper around rust's `Display<i64>` implementation.

```
~/G/j/jinko (int-to-string|✚2…) $ hyperfine --warmup 10 'python python_impl.py' 'target/release/jinko jinko_impl.jk' 'target/release/jinko native_rust_fmt.jk'
Benchmark #1: python python_impl.py
  Time (mean ± σ):      20.6 ms ±   1.3 ms    [User: 17.3 ms, System: 3.7 ms]
  Range (min … max):    14.4 ms …  23.1 ms    116 runs

Benchmark #2: target/release/jinko jinko_impl.jk
  Time (mean ± σ):      28.1 ms ±   1.7 ms    [User: 28.1 ms, System: 1.0 ms]
  Range (min … max):    20.5 ms …  30.7 ms    93 runs

Benchmark #3: target/release/jinko native_rust_fmt.jk
  Time (mean ± σ):       2.3 ms ±   0.9 ms    [User: 2.3 ms, System: 1.3 ms]
  Range (min … max):     0.0 ms …   5.5 ms    535 runs

  Warning: Command took less than 5 ms to complete. Results might be inaccurate.

Summary
  'target/release/jinko native_rust_fmt.jk' ran
    9.00 ± 3.57 times faster than 'python python_impl.py'
   12.26 ± 4.86 times faster than 'target/release/jinko jinko_impl.jk'
```

A side-effect of these fixes is that we also accidentally gained the power to
do generic type inference... But this behavior needs to be standardized and
tested on its own:

```rust
ext func debug[T](instance: T); // builtin function

debug(15);
debug("s");

type Complex(real: int, imaginary: int);

debug(Complex(real: 15, imaginary: 12));
```

No need to call `debug[int](15)` anymore!

- #474: You can now call generic methods properly
- #472: Code Inclusion was fixed!

**Full Changelog**: v0.3.0-jinx3...v0.3.0-jinx4

v0.3.0-jinx3 pre-release for jinko

Release v0.3.0-jinx2

This release uses the new parser from @SanderJSA

- Fix variable mutation bug #187

v0.2.0 release!

- Implementation of the typechecker
- Resolve method calls correctly, desugar them
- Scope-map improvements
- Code inclusion is now available
- Field access on custom types are available
- Various clean ups
- Add stdlib:
        - Add Maybe<int>
        - Add Range<int>
        - Add Iter<int>
- Remove `audit` blocks from the language
- Implement `for` loops

- Implement for loops

- Implement exit()