Use JDA's LRU MemberCachePolicy instead of caching everything at startup

[Copilot]

JDA already provides MemberCachePolicy.lru(n) and cache-aware CacheRestAction retrieval — no need to layer a custom Caffeine cache on top, or manually pre-check the JDA cache before calling retrieve*.

Changes

• MemberCachePolicy.lru(20_000) replaces MemberCachePolicy.ALL — JDA manages a bounded LRU eviction cache natively, capping memory without unbounded growth
• Simplified getMemberById / getUserById — removed manual getMemberById() / getUserById() pre-checks; retrieveMemberById() and retrieveUserById() return CacheRestAction which already hits the cache before making a REST call

// Before: redundant manual cache check + REST fallback
Member member = guild.getMemberById(userId);
if (member != null) return member;
return guild.retrieveMemberById(userId).complete();

// After: CacheRestAction handles cache + REST in one call
return guild.retrieveMemberById(userId).complete();