Fix GC stress log and analysis

[markples]

Summary: This includes multiple fixes to reenable the viewing of GC stress log events (from GC dprintf) in StressLogAnalyzer. It makes a small subset of the events available in all _DEBUG builds, which provides a build check in CI without overwhelming the stress log or runtime. This includes an audit of dprintf(1, ...) sites. More detailed GC stress logs can be enabled by changing the sources (as before).

Fixes to StressLogAnalyzer are least-effort because it is intended to be replaced by a C# version in #104999.

Details:

• gcDetailedStartMsg
  • Split gcDetailedStartMsg to fit under the 16-argument limit for the stress log but keep all arguments for SIMPLE_DPRINTF
  • Cleanup
• StressLogAnalyzer
  • Handle 16-argument messages
  • Look for a prefix for IS_GCSTART
  • Avoid buffer over-reads
  • Avoid crash on %5s
• TRACE_GC
  • Enable TRACE_GC in _DEBUG builds
  • Change default behavior to not (re)use upper bits of LogFacility. Leave the mechanism for private builds.
  • Audit dprintf(1) sites and change frequent ones to other values

Includes #108089 until it is merged
Fixes #108628

[markples]

"ILC: error IL1013: Error processing 'name'" is #108031

[Maoni0]

other than the few minor changes you took notes of, the rest LGTM!

[mangod9]

@Maoni0, is this good to merge or there are some changes still needed?

[Maoni0]

I see the typos in the 2 comments in gcpriv.h were still not fixed...

// HOST_64BIT is required because logging dprintf to the stress log us only
should be
// HOST_64BIT is required because logging dprintf to the stress log is only

// similarly on log messages. A dprintf(1) is generally called about once per
should be
// similarly on log messages. A dprintf(1) is generally called about a few times per

[mangod9]

Ok, thanks. @mrsharm, can you please look into merging this PR (looks like some conflicts need to be resolved and the comments updated per Maoni above). Thx!

[mrsharm]

@mangod9 / @Maoni0: addressed the merge conflict + the comment update. Do let me know if any more changes are needed.

[jkoritzinsky]

@mrsharm You've added back the native version of the StressLogAnalyzer. Can you apply the changes that were in conflict to the managed implementation and delete the native implementation? In particular the changes to the recognition of GCStart messages

[mrsharm]

Thanks for catching that, @jkoritzinsky! Removed the old files and updated the comment. And checked the build. Do let me know if any other changes are needed.