[FEA] Promote CAGRA from experimental

[cjnolet]

Tracking tasklist to promote CAGRA from experimental. Note that this is a subset of the tasks in #1392.

cc @tfeher @enp1s0 @teju85 @wphicks

• Shrink CAGRA binary size (CAGRA binary size too large. #1459)
• Compose CAGRA from RAFT primitives/helpers (Compose CAGRA from RAFT primitives/helpers #1604)
• Avoid unecessary device memory copies ([FEA] CAGRA avoid unnecessary device memory copies of dataset / knn graph #1479)
• Use rmm to fix stream ordering in prune (CAGRA use rmm allocator and fix stream ordering in prune #1461)
• Re-introduce template instantiations to improve compile times (Re-introduce CAGRA template instantiations to reduce compile time  #1443)
• int8/uint8 distance scaling (CAGRA int8 / uint8 distance scaling #1457)
• Add cagra to ANN benchmarks (Add RAFT ANN benchmark for CAGRA #1552)
• Add CAGRA gbench (Add CAGRA gbench #1496)
• Fix multi-CTA accuracy bug (CAGRA multi-cta solver accuracy #1575)

[enp1s0]

I'm now working on #1575.