fix(BACK-2167): reduce number of state generations

[Ivan011001]

Note

Medium Risk
Changes how FluidDex reserve data is sourced (cached polling vs per-request generation), so stale/missing cache entries can reduce pricing availability or accuracy until the poller runs.

Overview
Reduces on-demand FluidDex reserve state generation by replacing the FluidDexLiquidityProxy event-subscriber/stateful model with a cache-backed polling approach. A master instance now periodically fetches reserves on-chain (RESERVE_REFRESH_INTERVAL_MS = 10s) and writes them to shared cache (10 min TTL + 5s local TTL), while pricing/top-pools reads only from cache and returns null when state is unavailable.

Also increases Balancer V3 pool/rate refresh intervals from 1 to 2 minutes, updates FluidDex event tests’ reference block numbers, adds releaseResources() cleanup for the new polling interval, and bumps package version to 4.8.21.

^{Written by Cursor Bugbot for commit a03cdf5. This will update automatically on new commits. Configure here.}

[linear]

BACK-2167

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}