suggestion: create RedisAdapterLite

[d4vebr4cey]

Create a RedisAdapterLite that:

• Only provides necessary functionality (most RedisAdapter features are never used)
• Does not use templates
• Does not support clustering
• Only uses hiredis (probably eliminates tcache alignment error #73)
• Uses cmake in the typical ways the community does

Deprecate full RedisAdapter (or at least favor RedisAdapterLite)

[derekste]

Triage update:

origin/redis-adapter-lite is real and valuable, but it needs a clear route to main.

Current branch state:

• origin/redis-adapter-lite is 4 commits behind origin/main and 30 commits ahead.
• The branch replaces the old redis-plus-plus based implementation with a hiredis-based adapter, removes cluster support, adds a compatibility wrapper, and adds substantial tests/benchmarks/docs.
• It still includes pub/sub support, even though the GRAFE path appears to want a stream-only, no-cluster core.

Recommended path:

• Open a draft PR from redis-adapter-lite or rebuild it on current main so review can happen in GitHub instead of branch-only state.
• Decide explicitly whether pub/sub stays. If GRAFE does not use pub/sub, remove it from the lite path before presenting it as the core adapter direction.
• Split if needed: first land test/CI and bug-fix pieces, then the hiredis/lite implementation, then compatibility/adoption changes.

Related cleanup:

• refactor: eliminate cluster support #65 tracks removing cluster support.
• Remove unused pub/sub API from the core adapter surface #97 tracks the remaining pub/sub API decision.