Fix false unused import warning for given CanEqual in pattern matching

[Bbn08]

Problem

Fixes #25227

Under strict equality, import Givens.given (providing CanEqual[Any, Null]) is falsely reported as unused when the given is consumed only by pattern matching (case null =>), even though removing it breaks compilation. The warning does not appear when the given is consumed by ==/!=.

import scala.language.strictEquality

object Givens:
  given CanEqual[Any, Null] = CanEqual.derived

object Test:
  import Givens.given  // ← falsely reported as unused

  def m(value: Any): Boolean = value match
    case null => true   // needs CanEqual[Any, Null]
    case _ => false

Root Cause

CheckUnused.transformApply correctly resolves CanEqual for ==/!= via resolveScoped(caneq), marking the given import as used. However, transformMatch had no equivalent logic for pattern matching against literal/null patterns, so the import was never marked as used.

Fix

Added the same resolveScoped(caneq) logic to transformMatch for Literal patterns (including null). When a Match tree has a literal pattern, we construct CanEqual[SelectorType, PatternType] and call resolveScoped to mark any given import providing it as used.

Testing

New test tests/warn/i25227.scala with both positive (used by case null =>) and negative (genuinely unused) cases
All existing warn tests pass, including unused-can-equal.scala and i17762.scala

[som-snytt]

There is a PR #25114 to move the lint after patter matcher to detect these comparisons.

[Bbn08]

@som-snytt Thanks for pointing to #25114, I believe your fix should address this as well, since moving CheckUnused to run after pattern matching desugaring means literal/null patterns will already be lowered to equality checks that the existing transformApply logic handles. That's a much more elegant and comprehensive approach than what I had.

I had put together a smaller, targeted fix that adds resolveScoped(caneq) directly in transformMatch for literal patterns, which could potentially serve as a complementary defense-in-depth measure.

That said, if your merged fix already covers this case fully, I'm happy to close my PR please let me know. Thanks for the great work on this!

[som-snytt]

I confirmed it is fixed by the other PR, which is merged, but it would be nice to have this test without the code change!

[Bbn08]

I've updated this PR to be test only the code changes to CheckUnused have been removed, since #25114 (now merged) addresses the underlying issue by moving the lint phase after the pattern matcher.

What this PR now contains:

• A single new file tests/warn/i25227.scala, a regression test for False unused import warnings when importing "given" #25227
• The test verifies that -Wunused:all correctly reports unused given CanEqual imports under strictEquality, including the case where a given import is consumed only by case null => pattern matching

I'm sorry if this change took more than the expected time had something going on in life. Thank you.