[m3ninx] Avoid cloning of roaring bitmap in conjunctionSearcher

[linasm]

What this PR does / why we need it:
conjunctionSearcher.Search was always making a deep copy (Clone) of the first PostingsList, even though it was not necessary when there was more than one PostingsList in play. It was not necessary because both Intersect and Difference construct a new roaring bitmap under the hood (there are no in-place implementations for Intersect and Difference).

I have added a call to CloneAsMutable at the end of conjunctionSearcher.Search, it will only be executed if there were no Intersect/Difference calls made. I might have been too paranoid and could have avoided this because conjunctionSearcher.Search returns immutable List (so technically there would be no need to clone the input), but this all design does not fully embrace immutability (e.g. same object implements both mutable and immutable interfaces), which makes it very fragile, so I decided to err on the safe side.

In practice, I expect this clone in the end to almost never happen. Because even for a very basic Prometheus query of the form foo{bar="x"} you already have an intersection of 2 search predicates, meaning that the cloning gets avoided.

Note that the savings on heap size would not be very significant (I believe because those objects are mostly short lived):

But reduction of allocation/GC pressure in some use cases might be significant:

Special notes for your reviewer:
To make the immutable vs in-place operations more explicit, I have moved Intersect and Difference from MutableList to the immutable List interface and changed them to return the resulting value instead of doing a shallow mutation.
Also, renamed Union/UnionMany to UnionInPlace/UnionManyInPlace, to better reflect the semantics (and to match the names of the underlying bitmap methods used).
Also, renamed List.Clone to List.CloneAsMutable because it is returning a MutableList, so that it is not a proper clone method.

Does this PR introduce a user-facing and/or backwards incompatible change?:
NONE

Does this PR require updating code package or user-facing documentation?:
NONE

[codecov]

Codecov Report

Merging #3948 (f1ba705) into master (fc654f1) will decrease coverage by 0.2%.
The diff coverage is n/a.

❗ Current head f1ba705 differs from pull request most recent head 5a70a6e. Consider uploading reports for the commit 5a70a6e to get more accurate results

@@           Coverage Diff            @@
##           master   #3948     +/-   ##
========================================
- Coverage    56.9%   56.7%   -0.3%     
========================================
  Files         551     555      +4     
  Lines       63126   63406    +280     
========================================
+ Hits        35947   35975     +28     
- Misses      24003   24237    +234     
- Partials     3176    3194     +18     

Flag	Coverage Δ
aggregator	62.5% <ø> (-0.1%)	⬇️
cluster	∅ <ø> (∅)
collector	58.4% <ø> (?)
dbnode	60.4% <ø> (-0.3%)	⬇️
m3em	46.4% <ø> (ø)
metrics	19.7% <ø> (ø)
msg	74.1% <ø> (-0.1%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 2eace16...5a70a6e. Read the comment docs.

[Antanukas]

LGTM. Indeed cloning looks necessary if we do at least one intersection/union/difference operation.

[soundvibe]

LGTM from the code perspective, but don't have deep understanding of this codebase.