Optimise data transfer in TopologicalSorter

[jakelishman]

This optimises the creation of the list objects returned from TopologicalSorter.get_ready (by avoiding a temporary Vec) and allows done to accept a single index, avoiding the need for the user to allocate many small lists in several real-world uses.

This is the overload version of #1157. Happy to change it to a separate function if desired.

Close #1157

---

With a setup of:

from qiskit.circuit.library import EfficientSU2
from qiskit.converters import circuit_to_dag

import rustworkx

dag = circuit_to_dag(EfficientSU2(100, reps=5, flatten=True), copy_operations=False)._multi_graph

(I was using Qiskit 1.0.2 "ish", but it doesn't matter much) and a timing script:

sorter = rustworkx.TopologicalSorter(dag, check_cycle=False, check_args=False)
while ready := sorter.get_ready():
    for node in ready:
        sorter.done(node)

for this PR, and sorter.done([node]) (with a list) for main, I found main took ~510µs and this PR took ~275µs on my machine.

That's a common pattern for how we want to use this in Qiskit, though we'll only conditionally be calling done on some nodes, which is why the overhead of managing lists really hurts us.

[coveralls]

Pull Request Test Coverage Report for Build 8638783196

Details

• 54 of 58 (93.1%) changed or added relevant lines in 1 file are covered.
• 1 unchanged line in 1 file lost coverage.
• Overall coverage decreased (-0.02%) to 96.505%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
src/toposort.rs	54	58	93.1%

Files with Coverage Reduction	New Missed Lines	%
src/toposort.rs	1	92.24%

Totals
Change from base Build 8576567307:	-0.02%
Covered Lines:	17311
Relevant Lines:	17938

---

💛 - Coveralls

[enavarro51]

Just a couple of questions and then this looks about ready to go.

[enavarro51]

LGTM.