Decomp/populate gateset from backend toml

[yde773786]

This is a draft

Context:
Add a graph-decomposition pass based on the device toml backend.

Description of the Change:
Added @device-based-decomposition decorator to automatically run the graph-decomposition pass based on the device's target gateset. This requires running adjoint-lowering and ctrl-lowering pass first.

A DeviceBasedDecomposition pass is added that injects the three passes as required.

Tests in frontend for ensuring pass is added successfully through device preprocessing. MLIR tests added along with graph-decomposition MLIR tests, both in frontend, to have access to required PATH variables to run graph-decomposition.

Some minor issues in graph-decomposition are also addressed here.

1. Parsing target gateset when the target gateset itself has special characters such as ( which occur with Adjoint / Ctrl gates.
2. <n>C... -> C... in the OperatorNode.name. The id remains unaffected.
3. Missing integration tests. Need some clarification on this, due to the current test coverage appearing to be identical to rest of the device_preprocessing injected passes.

Benefits:

• User does not need to specify the target gateset to run graph-decomposition.

Possible Drawbacks:

• Currently, struggles with nested adjoint and ctrl regions. The test for the pass passes only because the order in which ctrl-lowering and adjoint-lowering` are added

Related GitHub Issues:

[sc-129198]

[kipawaa]

With respect to the adjoint- and control-lowering passes, there's some related discussion in #3216 .