Reduce locks for adhoc BF

[alonre24]

Describe the changes in the pull request

In ad-hoc BF mode of the tiered index, we call internally to getDistanceFrom of the flat and HNSW indexes. These calls access the indexes' data, and it is not safe to run insert/delete operations in parallel. Until now we were acquiring the locks internally, but since this is usually called for every vector individually, we see that the overhead of acquiring and releasing the locks is significant in that case.
Hence we introduce the new API for acquiring and releasing the locks for shared ownership and assume that the caller acquires the locks while calling the tiered index getDistanceFrom.

Mark if applicable

• This PR introduces API changes
• This PR introduces serialization changes

[codecov]

Codecov Report

Patch coverage: 92.30% and project coverage change: -0.12% ⚠️

Comparison is base (2d45add) 95.59% compared to head (9a7ff88) 95.48%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #419      +/-   ##
==========================================
- Coverage   95.59%   95.48%   -0.12%     
==========================================
  Files          65       65              
  Lines        4086     4093       +7     
==========================================
+ Hits         3906     3908       +2     
- Misses        180      185       +5     

Files Changed	Coverage Δ
src/VecSim/vec_sim_index.h	96.73% <0.00%> (-2.15%)	⬇️
src/VecSim/vec_sim_interface.h	100.00% <ø> (ø)
.../VecSim/algorithms/brute_force/brute_force_multi.h	87.36% <100.00%> (ø)
...VecSim/algorithms/brute_force/brute_force_single.h	89.47% <100.00%> (ø)
src/VecSim/algorithms/hnsw/hnsw.h	98.33% <100.00%> (-0.10%)	⬇️
src/VecSim/algorithms/hnsw/hnsw_multi.h	89.01% <100.00%> (-1.90%)	⬇️
src/VecSim/algorithms/hnsw/hnsw_single.h	90.66% <100.00%> (-0.59%)	⬇️
src/VecSim/algorithms/hnsw/hnsw_tiered.h	98.92% <100.00%> (+0.01%)	⬆️
src/VecSim/vec_sim.cpp	96.26% <100.00%> (+0.17%)	⬆️

... and 1 file with indirect coverage changes

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[meiravgri]

See comments below :)
The main suggestion is to include UnSafe in the API function's name so if we have an alternative safe solution in the future, we won't have to break the API again.
Also, since this can cause a bottleneck in writing requests, I think it is worth examining the option to first accumulate results up to some limit (on Redisearch side) and then calculate the accumulated vectors distances.