JProfiling fix to avoid too many internal pointers

[vijaysun-omr]

Jprofiling adds control flow to the IL as part of the lowering of the calls to profile values. This control flow splits blocks, and in the process creates temps for values commoned across the split point. If there are enough internal pointers commoned in the method, the lower may end up creating more than the max number of internal pointers allowed, thus causing the compile to fail. This commit keeps track of how many internal pointers are created and avoids lowering any more calls once we have reached a certain threshold. Also added an env var to control this threshold for experimenting in the future.

[vijaysun-omr]

@r30shah your review would be appreciated

[r30shah]

Changes looks good to me.

[r30shah]

Jenkins test sanity.functional,sanity.openjdk xlinux,plinux,zlinux,alinux jdk21

[JamesKingdon]

i ran this change with a testcase that was hitting the internal pointer limit and found that it resolved the problem.

[r30shah]

Here is the link of the PR build : https://openj9-jenkins.osuosl.org/job/PullRequest-OpenJ9/6901/

Looking at the failures,
S390x failures,

1. J9vmTest_4 : Same issue as J9vmTest_0 FindDeadlockTest Testing chained deadlocks Wrong deadlocked threads #16493
2. cmdLineTester_getPid_0 : cmdLineTester_getPid_0 RuntimeMXBean.getPID() returned 0 instead of #13085
3. testSCCMLTests1_openj9_1 : testSCCMLTests1_openj9 Add ROM classes to the existing shared cache with line numbers by using mprotect=all abort setExtraStartupHints #20953
4. testJitserverArguments_0 : testJitserverArguments, Test SSL success condition - failed to configure SSL ecdh #18599

Apart from those, I see cmdLineTester_callsitedbgddrext_openj9_0 and cmdLineTester_GCRegressionTests_1 failing with unable to find relevant existing issue and testJITServer_1 failed on both Linux on x86 for which I couldn't find matching existing issue (Though, a 5x grinder on x86 passes : https://openj9-jenkins.osuosl.org/job/Grinder/4085/

As change is making JProfilingValue optimization more conservative in bailing out the lowering the JProfilingValue calls when it hits the limit, it is highly unlikely to cause the above issue. For sanity I am going to launch another grinder with sanity.functional on x86 and LoZ before merging.

[r30shah]

Jenkins test sanity.functional xlinux,zlinux jdk21

[r30shah]

Failing test on s390b and x86 [1] linux are related to JIT server and seen on other builds as well : #21066.

Relaunching the test to get clean one

[1]. https://openj9-jenkins.osuosl.org/job/Test_openjdk21_j9_sanity.functional_x86-64_linux_Personal_testList_1/241/console

[r30shah]

Jenkins test sanity.functional xlinux jdk21

[r30shah]

sanity build for xlinux passes and zlinux failure was related to #21066 and other known issues which is not caused by changes in this PR. Merging the change based on amount of testing done.