[app] Runtime config tuning: php-fpm pool, FPM access log, opcache, redis eviction policy

[henzigo]

Context

The shipped runtime configuration carries legacy defaults worth revisiting:

• php-fpm pool undersized: pm.max_children = 20 caps concurrency at 20 requests
  per pod while the memory limit would fit ~60 workers; spikes ≥20 → 502s from nginx
• access.log = /dev/null: FPM access log (CPU time, memory peak per request) is
  thrown away; should go to stdout with a useful format + slowlog to stderr
• opcache interned_strings_buffer = 24 is small for a Symfony/Shopsys codebase
  (recommend 64) — memory_consumption review included; **** optional JIT for CLI/cron
• redis maxmemory-policy volatile-lru evicts only TTL'd keys — keys without TTL
  can push redis to OOM errors; decide volatile-lru (and guarantee TTLs app-side) vs allkeys-lru
• **** maxmemory 2200mb vs pod limit 2500Mi is intentional headroom — parametrize
  maxmemory from the values (derive from redis.resources.limits.memory) or document
• **** redis slowlog tuning (slowlog-log-slower-than 5000, slowlog-max-len 128)

Proposal

All of these live in values already (webserver.phpFpm.config, .opcacheConfig,
redis.config) — change the shipped defaults, with load-test/metrics validation for the
php-fpm pool sizing (correlates with the memory request change in #35).

Tasks

• php-fpm pool sizing (validated by load test on a reference project)
• FPM access log to stdout + slowlog
• opcache tuning (+ JIT decision for CLI)
• redis eviction policy decision + maxmemory parametrization + slowlog
• Golden snapshots, deviations entries

Related: #35