fix: preserve causality for future HLC updates

[efegokdemir]

Summary

Fix the HLC receive-event branch so an observed timestamp that is ahead of the local clock produces a timestamp greater than the observed value.

Changes

• Compare the merged logical time with the observed logical time when selecting the remote-ahead branch.
• Add a regression test using a future timestamp with a non-zero causal counter.

Testing

• git diff --cached --check — passed
• mvn -pl base-hlc -Dtest=HLCTest -DskipITs -Dcheckstyle.skip -Drat.skip -Dspotless.skip=true test — blocked by the available OpenJDK 27/Lombok compiler incompatibility (NoSuchFieldError); the repository targets Java 17 and no Java 17 runtime is installed

Fixes #285

[ImDanXie]

Verified independently — the fix matches the reference receive rule (Kulkarni et al.), and the added regression test is the exact repro from #285.

Since the PR notes the test could not be run locally (JDK/Lombok mismatch): we ran it against the compiled base-hlc classes (OpenJDK 17, same code) during the investigation behind #285 —

observed=(l=1790183330031,c=5)  ->  update() returned=(l=1790183330031,c=0)   # before: now < observed

and with this one-line change the assertion updated > observed passes (returns c = otherC + 1 = 6).

One note for whoever merges: the existing HLCTest.update() keeps feeding the clock its own get() output (l == l_m), which is why this branch was never covered before — worth keeping the new test pinned to a future l_m (as it is here) rather than a self-observed timestamp.