Allow caching subcircuits

[ohuettenhofer]

This adds the cached_subcircuit function which converts a QCircuit object to a SubcircuitGate object.
When this is added to another circuit, it is treated like a gate (unlike a normal circuit where all the gates are appended directly to the existing gates, and there's no subcircuit). This means that it only needs to be compiled once, and should also save some memory.

This is currently a very simple implementation that only caches the Tequila Compilation, nothing backend specific, but it can already speed up circuit Compilation significantly:

The following code takes me around 10 seconds to run:

import time
import tequila as tq
from tequila import CircuitCompiler
from tequila.circuit.gates import cached_subcircuit

subcircuit = tq.QCircuit()
for i in range(10):
    subcircuit += tq.gates.Ry(angle=i, target=0)

circuit = tq.QCircuit()
for i in range(100):
    circuit += subcircuit

start = time.time()
compiled = CircuitCompiler.error_correctable_gate_set(1e-6).compile_circuit(circuit)
print(f"Compilation time: {time.time() - start} s")

Replacing circuit += subcircuit with circuit += cached_subcircuit(subcircuit) decreases this time to ~0.7 seconds, while (in theory) being functionally equivalent.