Use greedy_topological_sort to sort operations in musical scores

[Acciaccatura]

substitutes nx.topological_generations for greedy_topological_sort when ordering operations in musical score layouts. additionally:

• creates appropriate topological generations taking into account the fact that allocs, or right-register operations in general, may reuse (and hence have a temporal dependence on) free operations.
• as the motivation for this change seems to be to match up with qubit counts (see Improve qubit counting logic to minimize qubit counts over multiple allocations #1636), adds a test case to ensure musical score data lines up with expected qubit count.
  • the implementation of the test case may depend on how the musical score generator formats its output, so I'm happy to find alternative ways to structure this test, or toss it altogether :^)

output ends up looking like this (though it should be noted this PR leaves the original display logic intact and still does not make use of topo_gen. i assume we would like to make use of that in a separate PR meant for cleaning up the musical scores?):

bloq = TestManyAlloc(5, False)
msd = get_musical_score_data(bloq.decompose_bloq())
fig, ax = draw_musical_score(msd)
fig.show()

fixes #1640

[gemini-code-assist]

Code Review

This pull request updates the musical score drawing logic in qualtran/drawing/musical_score.py to use a greedy topological sort and adds a test verifying that the musical score aligns with the qubit count. However, a critical bug was identified where the logic for updating topological generations and incrementing seq_x is incorrectly indented outside of the main loop, leading to rendering issues and incorrect topological generation assignments. A detailed code suggestion was provided to fix the indentation and logic.

[gemini-code-assist]

Code Review

This pull request updates the musical score drawing logic to use a greedy topological sort and track the depth of each qubit line to compute topological generations. It also adds a test to verify that the musical score aligns with the expected qubit count. Feedback on these changes includes addressing a potential KeyError when looking up newly allocated lines in y_to_topo_gen and removing a leftover assert False debugging statement in the new test.