fix: expose idleTimeout, timeToLive and validateAfterInactivity on HttpClient5 connection pool - #1293

Context

On BTP Cloud Foundry, NAT gateways silently drop idle TCP connections after a hyperscaler-specific timeout (GCP ~600s, Azure ~240s, AWS ~300s). Without connection pool eviction configured, the SDK keeps half-open sockets indefinitely. The next request that reuses one stalls for ~120s waiting for the OS TCP retry budget before the retry strategy reconnects.

This PR exposes three new opt-in configuration methods on ApacheHttpClient5FactoryBuilder so affected applications can configure connection pool eviction without changing behavior for other consumers:

• idleTimeout(Duration) — evict connections idle longer than this period
• timeToLive(Duration) — hard cap on total connection lifetime regardless of activity
• validateAfterInactivity(Duration) — probe connections after inactivity before reuse

All three default to null (no change in existing behavior). DefaultApacheHttpClient5Factory applies each only when explicitly set via a new buildConnectionConfig() helper.

Workaround for affected applications (usable today before a release):

ApacheHttpClient5Accessor.setHttpClientFactory(
    ApacheHttpClient5FactoryBuilder.create()
        .idleTimeout(Duration.ofSeconds(55))
        .timeToLive(Duration.ofMinutes(10))
        .validateAfterInactivity(Duration.ofSeconds(2))
        .build());

Feature scope:

• idleTimeout(Duration) added to ApacheHttpClient5FactoryBuilder
• timeToLive(Duration) added to ApacheHttpClient5FactoryBuilder
• validateAfterInactivity(Duration) added to ApacheHttpClient5FactoryBuilder
• DefaultApacheHttpClient5Factory applies each setting only when explicitly provided
• HC4 AbstractHttpClientFactory — follow-up PR (HC4 requires a background evictor thread)

Definition of Done

• Functionality scope stated & covered
• Tests cover the scope above
• Error handling created / updated & covered by the tests above
• Documentation updated
• Release notes updated