Add decomposition for gates.FSWAP

[renatomello]

Checklist:

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

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (4fbe5ae) 100.00% compared to head (104229f) 100.00%.
Report is 4 commits behind head on master.

❗ Current head 104229f differs from pull request most recent head cea634d. Consider uploading reports for the commit cea634d to get more accurate results

Additional details and impacted files

@@            Coverage Diff            @@
##            master     #1093   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files           63        63           
  Lines         8862      8860    -2     
=========================================
- Hits          8862      8860    -2     

Flag	Coverage Δ
unittests	100.00% <100.00%> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[BrunoLiegiBastonLiegi]

Thanks @renatomello, looks good.

[stavros11]

Thanks @renatomello. This is fine with me, just wanted to mention that now that we have the transpiler in qibo, we should consider moving the Gate.decompose method to the Unroller interface. The main idea behind this was that we can have different Unrollers for different applications. For example the FSWAP may be decomposed to a set of GPI2 and CZ for the hardware, but a different decomposition for some other application (if that makes sense and is useful).

That is of course if you and @BrunoLiegiBastonLiegi agree, but it is quite clear that the decompose does the same thing with the Unroller and I would rather focus on improving one approach than trying to maintain two competing ones. If you have suggestions to improve the transpiler these are also welcome.

[renatomello]

Thanks @renatomello. This is fine with me, just wanted to mention that now that we have the transpiler in qibo, we should consider moving the Gate.decompose method to the Unroller interface. The main idea behind this was that we can have different Unrollers for different applications. For example the FSWAP may be decomposed to a set of GPI2 and CZ for the hardware, but a different decomposition for some other application (if that makes sense and is useful).

That is of course if you and @BrunoLiegiBastonLiegi agree, but it is quite clear that the decompose does the same thing with the Unroller and I would rather focus on improving one approach than trying to maintain two competing ones. If you have suggestions to improve the transpiler these are also welcome.

Yeah I agree we should concentrate decompositions on the transpiler. This one is being done so @BrunoLiegiBastonLiegi can move forward with the Clifford simulator in #1076 and also it can serve as a benchmark for any unroller.