instantiate conditional types without a combined mapper for the cache lookup

[Max Schwenk (maschwenk)]

instantiateTypeWorker builds combineTypeMappers(t.mapper, m) for every conditional type it instantiates. getConditionalTypeInstantiation only uses that mapper to map the outer type parameters to the type arguments for the cache key, and on a miss it builds newTypeMapper(outerTypeParameters, typeArguments) from those. so the composite mapper is garbage right after the lookup either way

this adds getConditionalTypeInstantiationEx(t, m1, mapper, ...), which maps the type parameters the way CompositeTypeMapper.Map does (m1 first, then instantiate the result with mapper if m1 changed it, else mapper alone). getConditionalTypeInstantiation calls it with m1 == nil, so the other callers don't change. (not mapTypeWithCompositeMapper: that goes through getMappedType, which first replaces a distributed type parameter with its constraint, and CompositeTypeMapper.Map doesn't)

on our 38k-file program (37,943 files, 0 errors), median of 3, one process at a time:

	allocs	heap after check
main, single threaded	148.86M → 144.51M (−4.35M, −2.9%)	same
main, 4 checkers	246.89M → 238.12M (−8.8M, −3.6%)	same
main + #64475 + #64526 + 3 unopened follow ups, single threaded	148.42M → 144.08M (−4.34M, −2.9%)	same
main + #64475 + #64526 + 3 unopened follow ups, 4 checkers	247.09M → 238.41M (−8.7M, −3.5%)	same

allocs is Memory allocs from --extendedDiagnostics. the composite mappers are garbage right away, so this saves allocations and GC work, not retained memory: heap after check, symbols, types and instantiations don't change. an instrumented port of the checker counts 4,340,555 composite mappers avoided single threaded on that stack, and go's malloc count there goes down by 4,340,793. check time is within noise

same diagnostics on every run. go test ./... passes (except the macOS fsevents tests in internal/fswatch, which time out here under load and pass on a rerun), the testrunner also with TS_TEST_PROGRAM_SINGLE_THREADED=false, lint and format are clean. no new test, the change has no observable effect besides allocations

used claude code to help write this, ive reviewed it

[Max Schwenk (maschwenk)]

Withdrawing this one: the gain is too small to be worth reviewer time on its own. The larger change is #64475 / #64526.