Match grape infidelity with manually computed one

[rochisha0]

Fix #46
With this PR the GRAPE-reported infidelity matches the manually computed one by evolving the system using the optimized control pulses.
key change enable this consistency:

For state transfer problems, the correct fidelity normalization (scale_factor = 1/dim) is applied in the QOC interface, using existing optimization_type checks

[coveralls]

Pull Request Test Coverage Report for Build 16796973140

Details

• 7 of 14 (50.0%) changed or added relevant lines in 1 file are covered.
• 1 unchanged line in 1 file lost coverage.
• Overall coverage decreased (-0.5%) to 86.564%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
src/qutip_qoc/pulse_optim.py	7	14	50.0%

Files with Coverage Reduction	New Missed Lines	%
src/qutip_qoc/pulse_optim.py	1	80.69%

Totals
Change from base Build 16036109532:	-0.5%
Covered Lines:	844
Relevant Lines:	975

---

💛 - Coveralls

[rochisha0]

I’ve fixed the issue with looping over the objectives. The problem with using evo_full_final is that it returns the final evolution operator, not the evolved state. This matches what is done in qutip_qtrl/dynamics.py and is probably the reason why it is deprecated.

[ajgpitch]

The problem with using evo_full_final is that it returns the final evolution operator, not the evolved state

If we are doing state transfer then surely these are the same thing?

[ajgpitch]

This matches what is done in qutip_qtrl/dynamics.py and is probably the reason why it is deprecated.

What has been deprecated?

[ajgpitch]

Seems like a strange way to check if something is a density matrix

[ajgpitch]

In no way is a criticism of what @rochisha0 has done to fix the issue, but this all seems like a bit of a mess. Seems strange that we don't ask the user to be more specific about what their objective is rather than us trying to guess based on what they have input, especially if we have no elegant way to distinguish between a density matrix and unitary matrix say. Also, does this code allow for the initial / target to be a CPT map?

I am minded to merge this, as it seems to fix the issue, but open another issue that records the concerns raised above.

[rochisha0]

@ajgpitch , we are considering both state_transfer and gate_synthesis in this case.

[ajgpitch]

Okay, merging this now, despite concerns that this does not actually get tested on the assumption that #43 and #44 will be merged soon and will test this.

[pmenczel]

@ajgpitch Thanks for merging, I will go over those notebooks in #44 one more time and hope it will be ready for merging then.

Just to clarify though: this PR fixed the "symptom" but not the underlying problem of the final state being computed in multiple different ways, right?