Changed simulators fallback to decompose_once and removed ancilla support from DensityMatrixSimulator

[tanujkhattar]

See #6081 (comment) for a detailed description. Tl;Dr is as follows:

1. Fixed bugs in strat_act_on_from_apply_decompose and SimulationState which were introduced as part of Update Density Matrix and State Vector Simulators to work when an operation allocates new qubits as part of its decomposition #6108
2. Laid foundations for choosing a custom qubit allocation strategy when simulating circuits containing operations that can allocate new ancillas as part of their decompose protocol.

[NoureldinYosri]

The title of the PR doesn't really match the content of the PR, how about "Changed simulators' fallback to decompose_once and removed ancilla support from DensityMatrixSimulator".

does having DensityMatrixSimulator support ancillas block cirq-ft ?

---

nit: this PR is 2 in 1: the first a partial reversal of #6108 to remove the ancilla support from DensityMatrixSimulator and the second one contains the changes to strat_act_on_from_apply_decompose. so splitting it might be a good idea.

[tanujkhattar]

Updated the title.

does having DensityMatrixSimulator support ancillas block cirq-ft ?

No.

so splitting it might be a good idea

Hmm, I'll let people take a look at it as is for a bit and I can split if it's getting hard to review.

[NoureldinYosri]

LGTM. pending review from @senecameeks

[tanujkhattar]

@NoureldinYosri Can you LGTM with an approval to merge? I'll ping Seneca separately to take a look in the meantime.

[senecameeks]

Reviewed, looks good!

[NoureldinYosri]

please update the docstring of add_qubits in simulation_state before you merge