JIT: build pred lists before instrumentation

[AndyAyersMS]

Move pred list building to just before instrumentation (and just after importation -- we are getting very close to the front of the phase list now).

The block and edge count instrumenters were both using cheap preds to keep track of some relocated count probes. Revise this so they can use the regular pred lists. Also rework both approaches so their RelocateProbes methods are fairly similar and perhaps could be unified one day.

Contributes to #80193.

Diffs

[ghost]

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

Issue Details

---

Move pred list building to just before instrumentation (and just after importation -- we are getting very close to the front of the phase list now).

The block and edge count instrumenters were both using cheap preds to keep track of some relocated count probes. Revise this so they can use the regular pred lists. Also rework both approaches so their RelocateProbes methods are fairly similar and perhaps could be unified one day.

Contributes to #80193.

Author:	AndyAyersMS
Assignees:	AndyAyersMS
Labels:	area-CodeGen-coreclr
Milestone:	-

[AndyAyersMS]

@EgorBo PTAL
cc @dotnet/jit-contrib

For some context -- both block and edge count instrumentors were not only building cheap preds, but hacking the cheap pred lists to convey information from the RelocateProbes analysis to the actual instrumentation.

I obviously couldn't hack the regular pred lists the same way, so had to find different approaches. The result is simpler overall, especially on the edge probe side where the approach I added (just recently) was overly convoluted.

Some minimal diffs expected for optimized methods with edge probes, mainly because cheap preds did not have any ordering guarantees, and some of the edge instrumentation is sensitive to the order in which preds are enumerated.

[AndyAyersMS]

Failure is some kind of memory corruption on an arm test run. Almost certainly unrelated.

Assert failure(PID 23 [0x00000017], Thread: 44 [0x002c]): GCHeapUtilities::GetGCHeap()->IsInFrozenSegment(obj)
    File: /__w/1/s/src/coreclr/vm/syncblk.cpp Line: 2048
    Image: /root/helix/work/correlation/dotnet

[AndyAyersMS]

/azp run runtime-coreclr jitstress

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[AndyAyersMS]

jitstress failure is #81324