Change casting rules in various places

[mudit2812]

Lot's of places cast to float/complex (or don't cast at all when they should). This caused a huge splurge of warnings in many demos in qml (see https://github.com/PennyLaneAI/qml/issues/1212).

• Cast matrices to complex before applying to the state vector in devices.qubit.apply_operation. This should be fine since the state vector will always be complex anyway. This is also consistent with what the legacy devices do.
• Change how eigvals are cast to float for computing expvals. I don't entirely understand this, but it works.
• Stop casting global phases to complex when computing controlled decompositions. This is because recursively decomposing QROM will create controlled ops, and decomposing those will add controlled GlobalPhase with a real phase, but with a complex dtype.
• Cast theta to complex for all interfaces in SpecialUnitary.compute_matrix. This is needed because the matrix with which we tensordot theta is complex as it contains PauliY's matrix.

Open question: Do we want to do a bug fix release for this PR?

[sc-73429]

[github-actions[bot]]

Hello. You may have forgotten to update the changelog! Please edit doc/releases/changelog-dev.md with:

• A one-to-two sentence description of the change. You may include a small working example for new features.
• A link back to this PR.
• Your name (or GitHub username) in the contributors section.

[codecov[bot]]

Codecov Report

All modified and coverable lines are covered by tests :white_check_mark:

Project coverage is 99.63%. Comparing base (5f4169a) to head (84be34f). Report is 1 commits behind head on master.

Additional details and impacted files

```diff @@ Coverage Diff @@ ## master #6256 +/- ## ========================================== - Coverage 99.63% 99.63% -0.01% ========================================== Files 445 445 Lines 42433 42432 -1 ========================================== - Hits 42278 42277 -1 Misses 155 155 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

[mudit2812]

Looks like all tests are passing, but RTD has been building for an hour. Going to push a [skip ci] commit so tests don't run again, but RTD should still build.

[mudit2812]

Going to hold off on merging this as we might be cherry picking it into a bug fix release.

[mudit2812]

Closing in favour of #6260