[MRG] Use the weights w in ot.dist for all metrics

[raashish1601]

Types of changes

• Bug fix

Motivation and context / Related issue

ot.dist silently ignored w for every metric except cityblock and minkowski, so it returned the unweighted distance:

import numpy as np, ot
from scipy.spatial.distance import cdist
x, y, w = np.array([[0., 0.]]), np.array([[1., 1.]]), np.array([1., 2.])
ot.dist(x, y, metric="sqeuclidean", w=w)   # [[2.]]
cdist(x, y, metric="sqeuclidean", w=w)     # [[3.]]

This affected the backend implementations of sqeuclidean, euclidean, cosine and correlation, and the numpy fallback to cdist, which only passed w for minkowski/wminkowski. With backend="scipy" the weights were used, so the result depended on the backend. test_dist already lists these metrics as "those that support weights" but did not check the values. #859 fixed the same problem for cityblock.

Changes:

• sqeuclidean/euclidean: scale both point sets by sqrt(w) before euclidean_distances.
• cosine: same scaling. correlation: center with the weighted mean, then the same scaling. This matches the scipy definitions.
• numpy fallback: pass w to cdist whenever it is given (scipy still raises for metrics that do not accept weights).

Unweighted results are unchanged.

How has this been tested (if it applies)

Added test_dist_weighted_vs_cdist (numpy, compared with scipy.spatial.distance.cdist for sqeuclidean, euclidean, cosine, correlation, braycurtis and canberra) and test_dist_weighted_backends (all backends, for the four backend metrics). All new cases fail on master. pytest test/test_utils.py passes (numpy and torch backends installed locally), and ruff 0.5.2 lint and format are clean.

PR checklist

• I have read the CONTRIBUTING document.
• The documentation is up-to-date with the changes I made (check build artifacts).
• All tests passed, and additional code has been covered with new tests.
• I have added the PR and Issue fix to the RELEASES.md file.