Migrate from Guava Cache to Caffeine

[oneonestar]

Description

Migrate from Guava Cache to Caffeine.

Motivation:
#783 (comment)

Additional context and related issues

Removed the call to setRoutingGroupForQueryId() and setExternalUrlForQueryId() because cache.get() already insert the value into the cache if absent.

Release notes

(x) This is not user-visible or is docs only, and no release notes are required.

[Copilot]

Pull request overview

This PR migrates the caching implementation from Guava Cache to Caffeine, a modern caching library with better performance characteristics. The migration affects two main classes: TrinoRequestUser and BaseRoutingManager, where cache loading logic is simplified and exception handling is updated to catch RuntimeException instead of ExecutionException.

Key Changes:

• Replaced Guava Cache imports with Caffeine cache imports
• Simplified cache loading syntax using method references
• Updated exception handling from ExecutionException to RuntimeException

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
gateway-ha/src/main/java/io/trino/gateway/ha/router/TrinoRequestUser.java	Updated imports and simplified cache builder initialization with method reference
gateway-ha/src/main/java/io/trino/gateway/ha/router/BaseRoutingManager.java	Migrated cache implementation, removed redundant cache insertion calls, and simplified cache loader using method reference
gateway-ha/pom.xml	Added Caffeine dependency version 3.2.3

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Peiyingy]

LGTM! Thanks so much for making this change

[ebyhr]

In Trino, we intentionally avoid Caffeine (see trinodb/trino#10512 (comment)). Please hold off on this PR until we reach consensus.

[xkrogen]

@ebyhr I don't think the discussion from that issue is relevant here. GW cache doesn't have any invalidations to be concerned about. Even if we did have such concerns, GW is using the plain Guava cache, not Trino's EvictableCache, so (a) we are already not standardized, and (b) the behavior will at least be improved when moving from Guava to Caffeine.

I see two options here:

• Move to Caffeine -- This lets GW use the current best known in-memory Java caching solution, at the cost of some divergence from Trino (though no worse than the current situation).
• Move to EvictableCache -- trino-cache is already exposed as a separate module which we could pull in here. This at least unifies the implementation, so that any future improvements in Trino (e.g. moving the underlying implementation to Caffeine) will benefit GW as well.

Leaving things as-is only provides the illusion of standardization across Trino / GW ..