defer_measurements raises errors for unsupported measurements

[mudit2812]

Context:
Measurements that cover all wires can give users unexpected outputs when using mid-circuit measurements. So, defer_measurements now explicitly removes support for such measurements by raising errors.

Description of the Change:

• Raise an error if qml.state() is used.
• Raise an error if qml.probs(), qml.sample(), or qml.counts() is used if wires or an observable isn't specified.
• Add warning to docs

Benefits:

Possible Drawbacks:

Related GitHub Issues:
#4363

[albi3ro]

I know there's a lot of debate over this, but I'm fine with this decision and its better than people being confused about what their answers actually mean.

[mudit2812]

I know there's a lot of debate over this, but I'm fine with this decision and its better than people being confused about what their answers actually mean.

For reference, @trbromley and I had a discussion offline and the decision was to go in this direction in the end.

[Alex-Preciado]

@mudit2812 Thank you for validating with @trbromley and for posting the update here.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (a7114a9) 99.64% compared to head (a4cf441) 99.64%.
Report is 2 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #4701      +/-   ##
==========================================
- Coverage   99.64%   99.64%   -0.01%     
==========================================
  Files         379      380       +1     
  Lines       34126    33983     -143     
==========================================
- Hits        34005    33861     -144     
- Misses        121      122       +1     

Files	Coverage Δ
pennylane/_device.py	99.41% <100.00%> (+<0.01%)	⬆️
pennylane/devices/default_qubit.py	100.00% <100.00%> (ø)
pennylane/devices/qubit/sampling.py	100.00% <100.00%> (ø)
pennylane/devices/qubit/simulate.py	100.00% <100.00%> (ø)
pennylane/drawer/draw.py	100.00% <100.00%> (ø)
pennylane/math/single_dispatch.py	99.48% <100.00%> (+<0.01%)	⬆️
pennylane/measurements/mid_measure.py	100.00% <100.00%> (ø)
pennylane/operation.py	97.27% <100.00%> (-0.02%)	⬇️
pennylane/ops/op_math/composite.py	100.00% <100.00%> (ø)
pennylane/ops/op_math/controlled.py	99.63% <ø> (-0.01%)	⬇️
... and 7 more

... and 39 files with indirect coverage changes

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

[trbromley]

Looking back at this PR, we probably should have given a recommendation, i.e., if you're seeing an error because of qml.state(), try using qml.density_matrix() instead.