Bump bhtsne from 0.5.4 to 0.7 in linfa-tsne

[LucaCappelletti94]

Bumps bhtsne from 0.5.4 to 0.7 in linfa-tsne. The 0.6 release switched the embedding dimensionality from a builder parameter to a const generic tSNE<T, U, const D: usize>, so the wrapper now dispatches at runtime over D = 1, 2, 3 for exact t-SNE and D = 2, 3 for Barnes-Hut (the Morton<D> bound limits Barnes-Hut to those dimensions).

bhtsne 0.7 is considerably faster: the optimization loop is parallelized (roughly 4x per epoch at 2k points, scaling with N), the Barnes-Hut tree is now a Morton linear quadtree in a single contiguous arena (repulsive pass 2–3.6x faster, peak memory down by about a third), and there is an FFT-accelerated FIt-SNE fitting path available for future use. All existing tests pass.

[codecov]

Codecov Report

❌ Patch coverage is 80.64516% with 6 lines in your changes missing coverage. Please review.
✅ Project coverage is 77.87%. Comparing base (7fe5c86) to head (c93abe1).

Files with missing lines	Patch %	Lines
algorithms/linfa-tsne/src/lib.rs	80.64%	6 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #446      +/-   ##
==========================================
+ Coverage   77.53%   77.87%   +0.33%     
==========================================
  Files         106      104       -2     
  Lines        7585     7538      -47     
==========================================
- Hits         5881     5870      -11     
+ Misses       1704     1668      -36     

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[LucaCappelletti94]

Closing as I did not realize you had pinned 1.87.