LocalConnection1D/2D/3D: accept a given w and per-weight bounds - #840

What this changes and why

LocalConnection1D, LocalConnection2D and LocalConnection3D crashed in their constructors in two cases:

• A given w raised AttributeError: ... has no attribute 'out_channels'. The shape check used
  self.out_channels, which these classes never set; the random-w branch next to it uses self.n_filters.
  The check now uses self.n_filters.
• Per-weight wmin/wmax (a tensor the size of w, which AbstractConnection accepts) raised
  "Boolean value of Tensor with more than one value is ambiguous". The bound check was
  if self.wmin != -np.inf or self.wmax != np.inf, without .any(). It now has .any(), as in the other
  connection classes.

The docstrings of the three classes now say wmin/wmax may be a tensor, as the other classes' do.

Checklist

• I ran poetry run pytest locally and it passed. (653 passed, on a machine with CUDA)
• If I fixed a bug, this test fails on master and passes with my change
  (checked by the bug-test check):
  test/network/test_connections.py::test_local_connection_nd_given_weights_and_per_weight_bounds
  (given w, scalar and per-weight bounds, then 20 steps of PostPre learning; fails on master with
  the out_channels error) and
  test/network/test_connections.py::test_local_connection_nd_random_weights_within_per_weight_bounds
  (per-weight bounds without w; fails on master with the "ambiguous" error).
  One case per class for each. Locally the bug-test script reports Passed.
• I did not remove tests, remove checks, add skips or xfails, or widen tolerances.
  (weakened_tests.py locally: nothing found.)
• If I changed a formula: not applicable (a shape check and a condition).
• Does this change simulated results (spikes, voltages, traces, weights)?
  - [x] No. Both cases crashed before, so no working code changes.
• Did any of these files change? No.
• I added an entry to the Unreleased section of CHANGELOG.md.

AI assistance

• No AI tools were used.
• AI tools were used, and I (a human) have read every changed line, run the tests myself, and can explain the change in my own words.

Tools used and for which parts: Claude Code found both crashes while testing the GPU-bounds fix (#837),
wrote the fix, the tests and the changelog entry.