[MRG] speed up prefetch by adjusting calls to use len(minhash)

[ctb]

Several comparison dataclasses were introduced in #1955 and #2050, to facilitate gather/prefetch output and ANI calculation. Some of the methods in these dataclasses use len(minhash.hashes) which creates a new Python object and is typically quite slow.

This PR replaces len(minhash.hashes) in several key locations with len(minhash), which returns the same result but does so directly in Rust; on at least one benchmark, the speedup is about 10x (see numbers below).

Fixes #2124

timings

Using the benchmark from #1771 (comment) and also used here, we find:

This PR:

49.20 real        46.12 user         3.23 sys

latest branch / v4.4.2 -

469.38 real       419.82 user        47.47 sys

flamegraph profile for this PR

flamegraph profile for latest branch (equiv v4.4.2)

[codecov]

Codecov Report

Merging #2132 (0d94cde) into latest (2054391) will increase coverage by 7.38%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           latest    #2132      +/-   ##
==========================================
+ Coverage   84.31%   91.70%   +7.38%     
==========================================
  Files         130       99      -31     
  Lines       15293    11017    -4276     
  Branches     2167     2167              
==========================================
- Hits        12895    10103    -2792     
+ Misses       2095      611    -1484     
  Partials      303      303              

Flag	Coverage Δ
python	91.70% <100.00%> (ø)
rust	?

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

Impacted Files	Coverage Δ
src/sourmash/minhash.py	94.02% <100.00%> (ø)
src/sourmash/search.py	97.93% <100.00%> (ø)
src/core/src/ffi/nodegraph.rs
src/core/src/index/bigsi.rs
src/core/src/index/search.rs
src/core/src/cmd.rs
src/core/src/index/linear.rs
src/core/src/ffi/cmd/compute.rs
src/core/src/ffi/mod.rs
src/core/tests/test.rs
... and 23 more

[bluegenes]

Neat, did not realize that was the distinction between the two.

Looks great, thx!

[ctb]

Neat, did not realize that was the distinction between the two.

it took me a while but the flamegraph/benchmarking made it really clear!

(I knew that len(...) shouldn't be slow. Figuring out that it was slow was the important part ;)