Add custom decompose method to gates.U3 and gates.CZ

[renatomello]

Checklist:

• Reviewers confirm new code works as expected.
• Tests are passing.
• Coverage does not decrease.
• Documentation is updated.

[codecov]

Codecov Report

All modified lines are covered by tests ✅

Files	Coverage Δ
src/qibo/gates/gates.py	100.00% <100.00%> (ø)

📢 Thoughts on this report? Let us know!.

[alecandido]

Just a couple of comments about test and docs, but the feature on its own seems perfectly fine.

[alecandido]

@renatomello thanks to have applied the suggested corrections!

In general, it would be slightly better to pass a pRNG state, rather than using multiple seeds. It's because the properties of pRNG are optimized for sequential extractions, not necessarily for adjacent seeds.
However, this is certainly beyond the scope of this PR, since we should allow the various functions to get a state (i.e. a Generator object) instead of a seed.
And you should not notice any issue with just 10 extractions (these problems usually arise at the level of order of the generators' period extractions - or slightly less).

[stavros11]

Thanks @renatomello for adding this. Looks good to me.

Just noting that soon(ish) we will move the transpilers from qibolab to qibo and it would be great if these decompose methods are translated to transpilers to have a unified approach of gate transformations.

[renatomello]

Thanks @renatomello for adding this. Looks good to me.

Just noting that soon(ish) we will move the transpilers from qibolab to qibo and it would be great if these decompose methods are translated to transpilers to have a unified approach of gate transformations.

Could you open an issue as a reminder for this?