ForkJoinPool carrier starvation compensation for pinned virtual threads

[macarte]

When a pinned virtual thread's carrier blocks on a contended monitor, the ForkJoinPool has no visibility that its worker is blocked. If all carriers end up pinned on monitors, the VT holding the contested lock can never get a carrier to run on, causing deadlock.

This adds FJP compensation that activates an idle worker or creates a spare carrier when a pinned carrier is about to block on a monitor:

objectMonitor.cpp:

• compensate_pinned_carrier(): calls CarrierThread.beginMonitorBlock() before the carrier blocks, saving/restoring pending monitor state to handle nested monitor contention during worker thread creation
• end_compensate_pinned_carrier(): calls CarrierThread.endBlocking() after the carrier acquires the monitor
• New includes: symbolTable.hpp, javaCalls.hpp

ForkJoinPool.java:

• tryCompensateForMonitor(): variant of tryCompensate that omits the passive RC-only path (branch 2) and explicitly decrements RC so signalWork sees the pool as under-active while carriers are blocked
• beginMonitorCompensatedBlock(): spin-loop wrapper returning the RC_UNIT restoration value for endCompensatedBlock
• endCompensatedBlock: updated javadoc for monitor compensation path

JavaUtilConcurrentFJPAccess.java:

• Added beginMonitorCompensatedBlock to the shared secret interface

CarrierThread.java:

• beginMonitorBlock(): re-entrancy-guarded compensation that pins the continuation during FJP compensation to prevent preemption
• ForkJoinPools.beginMonitorCompensatedBlock(): static bridge method

[macarte]

manual rerun results

Here's the PR4 rerun summary (run 22935801854): 10 failures across 85 jobs

  Windows-aarch64 failures (3 jobs, 5 errors):

  ┌──────────────────────────┬─────────────────────────────────────────┬──────────────────────────────────────┐
  │ Job                      │ Test                                    │ Error                                │
  ├──────────────────────────┼─────────────────────────────────────────┼──────────────────────────────────────┤
  │ jdk/tier1 part 1         │ PingPong.java#ltq                       │ Timeout 480s (stuck at 2447/500000)  │
  ├──────────────────────────┼─────────────────────────────────────────┼──────────────────────────────────────┤
  │                          │ PingPong.java#sq                        │ Timeout 480s (stuck at 10875/500000) │
  ├──────────────────────────┼─────────────────────────────────────────┼──────────────────────────────────────┤
  │                          │ Skynet.java#default                     │ Timeout 6400s                        │
  ├──────────────────────────┼─────────────────────────────────────────┼──────────────────────────────────────┤
  │ hs/tier1 serviceability  │ GetStackTraceSuspendedStressTest.java   │ Timeout 480s                         │
  ├──────────────────────────┼─────────────────────────────────────────┼──────────────────────────────────────┤
  │ hs/tier1 compiler part 3 │ TestSpeculationFailedHigherEqual.java   │ Failed                               │
  └──────────────────────────┴─────────────────────────────────────────┴──────────────────────────────────────┘

  Cross-platform failures (7 jobs — all pre-existing):

  ┌─────────────────────────────────────────┬─────────────────────────────────────────────────────────────────────────────┐
  │ Test                                    │ Platforms                                                                   │
  ├─────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────┤
  │ TestSpeculationFailedHigherEqual.java   │ All 5 platforms (linux-x64, linux-x64-static, macos-aarch64, windows-x64,   │
  │                                         │ windows-aarch64)                                                            │
  ├─────────────────────────────────────────┼─────────────────────────────────────────────────────────────────────────────┤
  │ TestSharedCloseJFR.java                 │ linux-x64, linux-x64-static, macos-aarch64                                  │
  └─────────────────────────────────────────┴─────────────────────────────────────────────────────────────────────────────┘

  Key observations:

   - Same win-aarch64 virtual thread timeouts as PR2/PR3 — expected since PR4 doesn't add Dekker fences (that's PR5)
   - TestSpeculationFailedHigherEqual.java continues failing on all platforms — this is a tip-wide regression, not related 
  to your changes
   - TestSharedCloseJFR.java is the same persistent flaky test
   - jdk/tier1 part 2 passed this time (no SPMC timeout)
   - The GetStackTraceSuspendedStressTest timeout reappeared (was seen in PR #51 too)