test_coinmoves.py::test_coinmoves_unilateral_htlc_timeout: use pytest.approx for change amount

[whitslack]

Change amount can vary slightly. Exact comparisons cause test flakiness.

Sometimes I get an AssertionError because the change output is 15586 sat instead of the expected 15579 sat. Maybe this is due to a shorter-than-usual signature encoding. Use pytest.approx to allow the actual change amount to be within 7 sat of 15579 sat. (If the discrepancy is indeed due to a short signature encoding, then this could still be too strict, with a probability of 1 in 256, as theoretically a signature component could have a value arbitrarily close to zero.)

Checklist

Before submitting the PR, ensure the following tasks are completed. If an item is not applicable to your PR, please mark it as checked:

• The changelog has been updated in the relevant commit(s) according to the guidelines.
• Tests have been added or modified to reflect the changes.
• Documentation has been reviewed and updated as needed.
• Related issues have been listed and linked, including any that this PR closes. None found.

[rustyrussell]

But we are supposed to detect short sigs and skip this test! I've been meaning to diagnose why that doesn't work...

[whitslack]

Something has been bugging me about this. Isn't there a chicken-or-egg situation here? How can the value of a change output vary depending on the length of a signature encoding if the signature covers that change output? Is the signature here a SIGHASH_NONE or SIGHASH_SINGLE rather than a SIGHASH_ALL, such that the signature can be computed before we've decided on the change amount?