feat: Constructor generic type inference

[mattjohnsonpint]

Fixes #2893.

Changes proposed in this pull request:

Enables type inference for constructors of generic classes by:

• Refactoring to extract the existing type inference logic for functions that was in maybeInferCall, into its own private method inferGenericTypeArguments.

• Calling that new method from resolveClassInclTypeArguments when a generic class's constructor is called without any type arguments.

• Expanding the unit tests to verify working functionality and no regressions.

• I've read the contributing guidelines

• I've added my name and email to the NOTICE file

[mattjohnsonpint]

Note - I didn't change any of the actual existing logic for type inference that was used for functions. I just extracted it for re-use by constructor "new expressions". The PR is best reviewed with the "hide whitespace" box checked.

[CountBleck]

@mattjohnsonpint Thanks, that diff looks so much more manageable to review now.

[CountBleck]

Note: it doesn't handle this case:

class Foo<T> {
    constructor(public x: T) {}
}

class Bar<T> extends Foo<T> {}

const x = new Bar("hi")

That can probably be figured out at a later date.

[CountBleck]

This looks good, probably.

[mattjohnsonpint]

Note: it doesn't handle this case:

class Foo<T> {
    constructor(public x: T) {}
}

class Bar<T> extends Foo<T> {}

const x = new Bar("hi")

That can probably be figured out at a later date.

I might be able to address that. I'll take a look tomorrow. Good catch!

[mattjohnsonpint]

It should handle that case now. Thanks!

[mattjohnsonpint]

Also, FWIW, I don't think either of the previously existing test cases marked TODO are valid. They aren't in regular TypeScript anyway, so I don't think they should be in AssemblyScript.

[CountBleck]

This seems good.

[CountBleck]

They aren't in regular TypeScript anyway, so I don't think they should be in AssemblyScript.

I don't know what those are for either. I'm guessing it's something related to omitting : T and inferring the return type (I think a is a placeholder for some value dependent on the generic T).