Add simplify of U3->U1/U2 in OneQubitEulerDecomposer

[enavarro51]

Summary

Fixes #5468
Fixes #5556

Details and comments

This PR will simplify U3 gates in the OneQubitEulerDecomposer to U1 when theta is 0 and to U2 when theta is pi/2. This will allow the use of Optimize1qGatesDecomposition exclusively or in additional cases over Optimize1qGates. This first draft leaves the optimization as is.

The next phase will be to implement the conditions for optimization as discussed in #5531.

[enavarro51]

@kdk and @mtreinish. The latest commits fix several issues with Optimize1qGatesDecomposition (O1GD). See #5556. The solution here was to skip over runs with delay, snapshot, or custom gates. Alternatively these checks could be added to the collect_1q_runs filter_fn function.
Edited: Actually it looks like #5570 will take care of this problem. This draft includes that change for testing purposes assuming #5570 will be merged first.

In the current version of this draft, setting optimization_level=1 will implement the previous behavior where Optimize1qGates (O1G) is called if U1, U2, or U3 is in the basis. Setting optimization_level at 2 or 3 will use O1GD exclusively. This would allow for some performance and other testing to see how O1GD works by itself.

The final step in this PR would be to implement Matthew's idea for choosing between O1G and O1GD from #5531.

[kdk]

Thanks @enavarro51 , this looks good. One comment below. Also, can you add a release note?

[kdk]

Thanks for the updates @enavarro51 , this is looking good. Few comments below.

[kdk]

Thanks @enavarro51 , this is looking good. Two small comments below and then I think this is good to go.