feat(nearest): add IntervalTree.nearest()

[SahilRajputdotexe]

This adds IntervalTree.nearest(target, k=None), which finds the interval closest to a point or an Interval. Closes #114.

With k=None it returns one Interval, or None if the tree is empty. With an integer k it returns a list of up to k intervals, nearest first. Distance is the existing Interval.distance_to(): 0 when the two overlap or touch, otherwise the size of the gap. Ties are broken by the normal Interval ordering, so the result is stable even when the data fields can't be compared with each other.

>>> t = IntervalTree.from_tuples([(0, 5, 'a'), (10, 15, 'b'), (30, 40, 'c')])
>>> t.nearest(8)
Interval(10, 15, 'b')
>>> t.nearest(Interval(20, 22), k=2)
[Interval(10, 15, 'b'), Interval(30, 40, 'c')]

It doesn't scan the whole tree. It looks at a window around the target, using the boundary table to include intervals exactly at the window's edge, and doubles the window until it holds k intervals. A null Interval target raises ValueError, a negative k raises ValueError, and a non-integer k raises TypeError.

I added tests in test/intervaltree_methods/nearest_test.py, including a randomized comparison against a brute-force sort, plus README and CHANGELOG entries. The suite passes locally on Python 3.9, 3.12 and 3.14, and the full CI matrix (2.7, 3.5 to 3.14, flake8) passes on my fork.

I kept this to nearest only. A custom distance callable, a "within distance" query and cross-tree joins could be built on top of it if you want them, but I didn't want to widen the API in one PR.