Add indexing of NLayout

[jakelishman]

This allows NLayout to be indexed by VirtualQubit and PhysicalQubit, which is a shorter way of writing VirtualQubit::to_phys and PhysicalQubit::to_virt.

[qiskit-bot]

One or more of the following people are relevant to this code:

• @Qiskit/terra-core

[coveralls]

Pull Request Test Coverage Report for Build 16969118917

Details

• 4 of 10 (40.0%) changed or added relevant lines in 3 files are covered.
• 19 unchanged lines in 5 files lost coverage.
• Overall coverage decreased (-0.01%) to 88.247%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
crates/circuit/src/nlayout.rs	0	6	0.0%

Files with Coverage Reduction	New Missed Lines	%
crates/circuit/src/parameter/parameter_expression.rs	1	83.39%
crates/transpiler/src/passes/sabre/dag.rs	1	96.97%
crates/qasm2/src/lex.rs	3	92.53%
crates/qasm2/src/parse.rs	6	97.09%
crates/circuit/src/parameter/symbol_expr.rs	8	74.5%

Totals
Change from base Build 16967470304:	-0.01%
Covered Lines:	87840
Relevant Lines:	99539

---

💛 - Coveralls

[debasmita2102]

Just a naive question, in add_heuristic_layouts the dense trial now uses problem.dag.num_qubits() instead of the hardware width (num_physical_qubits) when calling compute_dense_starting_layout, so dummy ancillas are no longer included in this initial layout. Are we documenting any concrete changes in Sabre’s behaviour from this—either in runtime or in SWAP/depth quality—on the benchmark circuits, particularly for cases with many unused ancillas versus circuits that already occupy most of the device? Is that needed?

[jakelishman]

This was previously dependent on #14605 for no particular reason, but I've uncoupled it to allow the Sabre-layers chain to move forwards, since #14605 is slightly contentious given the discussion of #14917.

It's now unblocked and can proceed.

[jakelishman]

Just a naive question, in add_heuristic_layouts the dense trial now uses problem.dag.num_qubits() instead of the hardware width (num_physical_qubits) when calling compute_dense_starting_layout, so dummy ancillas are no longer included in this initial layout. Are we documenting any concrete changes in Sabre’s behaviour from this—either in runtime or in SWAP/depth quality—on the benchmark circuits, particularly for cases with many unused ancillas versus circuits that already occupy most of the device? Is that needed?

Oh, I never saw this at the time, sorry. If that change was showing up in this PR, it must have been because it was in a chain before. The PR that made the change you're talking about is #14604, which did come with a release note.

[alexanderivrii]

Neat feature that improves ergonomics - LGTM!