【代码贡献】Add HHL solver for linear systems (Ax=b)

[mnn31]

Implements HHL for solving Ax = b. The package init already had a
commented-out # from . import HHL slot reserved, so this just fills it in.
Relates to #13.

HHL(A, b).run() returns the normalized solution for Hermitian A. Standard
pipeline: amplitude-encode b, QPE on A (Oracle(e^{iAt}) for the controlled
evolutions, the existing plugin.QFT for the inverse), controlled RY eigenvalue
inversion on an ancilla rotating by 2*arcsin(C/lambda), inverse QPE to uncompute
the clock, then post-select the ancilla on |1>. t and C default to values from
A's spectrum but you can override both.

Checked solutions against numpy.linalg.solve. Fidelity is 1.0 on the diagonal
cases and the well-conditioned 2x2s; the [[3,1],[1,2]] case sits at ~0.9999
because its eigenvalues don't land exactly on the clock grid at 4 qubits, which
is the usual HHL resolution thing and improves with more clock qubits. Noted
that in the docstring.

Tests in Test_HHL_HHL.py cover the above plus input validation (non-Hermitian
and non-power-of-2 sizes raise ValueError). There's a doctest on the class too.

One limitation: A has to be Hermitian for now. Non-Hermitian works via the usual
dilation trick but I left that out to keep the diff small.