Allow VP to remove/upgrade virtual guards with merged HCR/OSR guards

[jdmpapin]

Allow folding (or for profiled guards, replacement by noppable guards) to proceed even if the guard in question is merged with an HCR or an OSR guard. Guards folded to always go to the cold side need no special handling. Guards folded to always go to the hot side (i.e. to fall through) have any merged HCR and/or OSR guard split out so that the virtual guard can be removed. When a profiled guard is replaced by a noppable guard, the noppable guard has the same merged HCR or OSR guard as the original profiled guard did, if any.

Currently, this only affects guards that can be folded simply based on the value numbers or constraints of the children (e.g. mustBeEqual()), or guards that use VFT test or method test. The logic that folds nonoverridden guards has canFoldNonOverriddenGuard() check separately for canBeRemoved(). In principle it should be fine to relax that check too, but canFoldNonOverriddenGuard() is also used in constrainCall(), which instead of using removeConditionalBranch() has its own way of removing the guard. And it needs its own way. For example, it has to do TR::TransformUtil::removeTree(...), not vp->_curTree->setNode(NULL). It also might not be able to setUnreachablePath() on the edge. So to extend this improvement to the elimination of nonoverridden guards, some of this new logic will need to be factored out into its own function, and that work is left for the future.

[jdmpapin]

The Mac failure is #6516:

2022-10-12T15:28:03.8651940Z 31: [----------] 19 tests from PortSockTest
...
2022-10-12T15:28:04.9457800Z 31: /Users/runner/work/1/s/fvtest/porttest/omrsockTest.cpp:1209: Failure
2022-10-12T15:28:04.9460540Z 31:       Expected: rc
2022-10-12T15:28:04.9469620Z 31:       Which is: 1
2022-10-12T15:28:04.9484140Z 31: To be equal to: 2
2022-10-12T15:28:04.9485900Z 31: [  FAILED  ] PortSockTest.poll_functionality_basic (1076 ms)
2022-10-12T15:28:04.9487750Z 31: /Users/runner/work/1/s/fvtest/porttest/omrsockTest.cpp:51: Failure
2022-10-12T15:28:04.9490820Z 31:       Expected: privateOmrPortLibrary->sock_bind(privateOmrPortLibrary, *serverSocket, serverSockAddr)
2022-10-12T15:28:04.9493200Z 31:       Which is: -506
2022-10-12T15:28:04.9495310Z 31: To be equal to: 0
2022-10-12T15:28:04.9497060Z 31: /Users/runner/work/1/s/fvtest/porttest/omrsockTest.cpp:1259: Failure
2022-10-12T15:28:04.9499890Z 31:       Expected: privateOmrPortLibrary->sock_accept(privateOmrPortLibrary, serverSocket, &connectedServerSockAddr, &connectedServerSocket)
2022-10-12T15:28:04.9502460Z 31:       Which is: -20
2022-10-12T15:28:04.9504100Z 31: To be equal to: 0
2022-10-12T15:28:04.9505750Z 31: [  FAILED  ] PortSockTest.poll_functionality_many_sockets (1 ms)
2022-10-12T15:28:04.9508210Z 31: [----------] 19 tests from PortSockTest (1082 ms total)
2022-10-12T15:28:04.9509780Z 31: 
2022-10-12T15:28:04.9511550Z 31: [==========] 245 tests from 21 test cases ran. (122829 ms total)
2022-10-12T15:28:04.9513200Z 31: [  PASSED  ] 243 tests.
2022-10-12T15:28:04.9514740Z 31: [  FAILED  ] 2 tests, listed below:
2022-10-12T15:28:04.9516400Z 31: [  FAILED  ] PortSockTest.poll_functionality_basic
2022-10-12T15:28:04.9518220Z 31: [  FAILED  ] PortSockTest.poll_functionality_many_sockets

[jdmpapin]

@vijaysun-omr, would you mind reviewing?

[vijaysun-omr]

Yes I will look at it this afternoon

[vijaysun-omr]

I tried thinking of a case when the class that we use (from the inlined method) as the one to check redefinition for in the HCR guard may be the wrong one, but came to the conclusion that the method must be defined in the class that is mentioned on the call node or a superclass. In either case, it seemed like we would have the class mentioned in the call node be redefined (i.e. it is redefined itself OR a superclass is redefined but that should result in the subclasses getting redefined too). Since this is the core change in logic, I think it is good.

[vijaysun-omr]

Jenkins build all

[jdmpapin]

The PPC64LE Linux failure is #6571:

[2022-10-14T00:03:56.123Z] 30: [----------] 14 tests from Special/PPCDirectEncodingTest
[2022-10-14T00:03:56.123Z] 30: free(): invalid next size (normal)
[2022-10-14T00:03:56.123Z] 30/30 Test #30: compunittest ......................Child aborted***Exception:   0.80 sec
...
[2022-10-14T00:03:56.123Z] The following tests FAILED:
[2022-10-14T00:03:56.123Z] 	 30 - compunittest (Child aborted)
[2022-10-14T00:03:56.123Z] Errors while running CTest

[vijaysun-omr]

Failing checks are known/unrelated (as per comments) and review is done. Merging.