Feat/acquire with timeout

[battermann]

• We added timeout-aware pool acquisition to hasql-resource-pool as a separate API entry point instead of changing the existing use behavior.
• connAcqTimeout keeps its original meaning: it configures libpq connect_timeout for connection establishment. It is not reused for pool-slot waiting, to avoid conflating two different timeouts.
• The new API exposes an explicit pool acquisition timeout parameter so callers can decide per call whether checkout should be bounded or unbounded.
• We kept the existing use / useWithObserver behavior unchanged and added timeout-aware variants for callers that need bounded checkout.
• We first tried a simple timeout (ResourcePool.takeResource pool) approach, but it did not reliably enforce the timeout in exhausted-pool scenarios.
• The final implementation mirrors resource-pool’s checkout mechanics and races resource availability against a timeout, which matches the semantics we needed more closely.
• The implementation is intentionally lower-level because resource-pool does not expose a public “bounded checkout” API.
• The test was rewritten to avoid relying on a slow SQL query holding the slot. It now blocks the first checkout in the connection getter itself, then verifies that a second checkout returns AcquisitionTimeoutUsageError while the pool is exhausted.
• This keeps the behavior deterministic and tests the acquisition timeout directly rather than depending on query execution timing.