Fix lambda normalization by renumbering parameters from innermost scope

[rishi-jat]

/claim #379

Normalize lambda parameter depths in LightTypeTagRef.Lambda
by recursively renumbering parameters from the innermost scope outward.

Normalization is applied locally during lambda construction and avoids
any changes to global parameter numbering or application logic.

Implements the approach suggested by @pshirshov.

Closes #379

[Copilot]

Pull request overview

This PR refactors lambda normalization in LightTypeTagRef by removing the FAKE parameter substitution approach and implementing a depth-based renumbering scheme. The new approach assigns depth 0 to innermost lambda parameters and negative depths (-1, -2, etc.) to outer scopes, making the normalized form suitable as the primary representation.

Key changes:

• Replaces makeFakeParams and lambdaFakeParamDepth with a recursive normalize function
• Implements depth renumbering based on lambda nesting level
• Removes normalizedParams as it's no longer needed with the new approach

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[rishi-jat]

@pshirshov @neko-kai , PR is ready for review please have a look when you get a chance. Thanks!

[neko-kai]

You should at least add the test from the issue itself to check the new normalized forms

[rishi-jat]

@neko-kai please have a look when you get a chance. Thanks!

[rishi-jat]

@pshirshov @neko-kai, can you please review the PR ? Thanks!

[rishi-jat]

/cc @pshirshov
/cc @neko-kai

[pshirshov]

Could you add some tests for string representations of the renormalized lambdas?

[rishi-jat]

Could you add some tests for string representations of the renormalized lambdas?

will do asap. Thanks!

[rishi-jat]

hi @pshirshov @neko-kai ,
I’ve added a regression test that checks the string representation of normalized lambdas.

The test now covers a real nested lambda case, verifies that FAKE_* parameters are gone, and asserts that outer-scope references are rendered with the correct depth in the output. All tests pass on Scala 2.12, 2.13, and 3.

Please let me know if you’d like the assertion to be stricter or if there’s another case you want covered.

[rishi-jat]

/cc @neko-kai
/cc @pshirshov

[rishi-jat]

@neko-kai @pshirshov , did you get a chance to review the PR ? Thanks!