cutqc2 run command with optimizations

[vineetbansal]

Major changes

• Introduced a cutqc2 run command (initially cutqc2 cut did an implicit run as well). This step is MPI capable.
• Documentation needs to be updated to reflect this!
• Removed multithreading for run_subcircuits method in CutCircuit class (multi-threading interferes with GPU execution and actually slows things down. MPI is a more scalable approach).
• Optimizations in implementation of Eq 3 - vectorized approach for binning.
• GPU backend for AerSimulator is now used by default. This is configurable through config.yaml as well as environment variables. This may cause the CI to fail on Mac since it is enabled by default. This may need some tweaking in the CI workflows as well as a mention in the documentation. Additional dependencies have been introduced requirements.txt for this.

EDIT: All the TODOs above have been attended to.

Overall, a 56x speedup even if just using a single MPI node.

[vineetbansal]

@Subashkatel - I'm done with this PR. Feel free to review and merge.

[Subashkatel]

committed