JIT: fix regressions from inlining policy change

[AndyAyersMS]

Don't boost the IL size threshold at Tier1+Instr, since this can cause us to lose profile data for key inlinees.

Fixes most of the regressions from #115904

[Copilot]

Pull Request Overview

This PR refines the inlining policy by disabling the IL size boost when running at Tier1 with instrumentation, preventing loss of profile data for critical inlinees.

• Added a check for opts.IsInstrumentedAndOptimized() to skip raising maxCodeSize in Tier1+Instr modes.
• Updated the JITDUMP message to reflect when the boost is suppressed.

Comments suppressed due to low confidence (1)

src/coreclr/jit/inlinepolicy.cpp:1386

• Add or update unit tests for the instrumented+optimized scenario to verify that maxCodeSize is not boosted as intended.

                if (!isTier1Instr)

[dotnet-policy-service]

Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
See info in area-owners.md if you want to be subscribed.

[AndyAyersMS]

@EgorBo PTAL
cc @dotnet/jit-contrib

Tier1+Instr has to balance perf vs degree of instrumentation: If we inline more we get better Tier1+Instr perf, but we may lose Tier1 perf. This PR returns us to the level we were at before #115119 and should fix many of the regressions reported in #115904.

Arguably we should be even less aggressive in Tier1+Instr (one way to see this is to look at the gap in perf between default and R2R disabled).

Someday we'll fix #44372 and we can revisit all this with fewer compromises.

[AndyAyersMS]

@EgorBot --filter "System.Globalization.Tests.StringEquality.Compare_DifferentFirstChar*"