Remove custom transpilation passes

[bachase]

As part of #246, the team benchmarked the impact of the customized transpilation passes compared with the version of the equivalent passes in the latest version of Qiskit. The results show the customizations are not currently providing additional gate savings or compilation time improvements compared with the native Qiskit passes.

This PR removes these customized passes and replaces them with the Qiskit equivalents in the default set in UCCDefault1.

[jordandsullivan]

I just manually triggered a benchmark run to check performance, and it looks good: same gate reduction and faster runtime. Although we are no longer focused on the narrative of being "the best compiler," this change actually brings us to be the top performing compiler in terms of speed and compiled ratio for our benchmarks :)
https://github.com/unitaryfund/ucc/blob/6c2c00ca9a7680275decda2dce30878241accd18/benchmarks/avg_compiler_benchmarks_over_time.png

Also, perhaps we should consider adding the Run benchmarks step as an automatic step in PRs -- only reason I didn't was that I didn't want them to run every time code was pushed to an open PR, plus it can create annoying merge conflicts when replacing the .pngs.

[jordandsullivan]

Thanks for doing this investigation and finding that we can get this speedup by using the latest Qiskit versions of our UCC pass set.

[bachase]

Thanks!

I agree on having these run as part of PRs. I think I've got a suggestion on how to do it without causing the conflicts, but need to write it down to validate in #235. Hoping to get my thoughts in markdown this week/next at the latest.