[ Feature ] Post-selected measurement #239

Given a generalized tableau, there is an existing feature

GeneralizedTableau.measure(index:int)->MeasurementResult

that lets one measure a qubit and project/mutate it to a new state given the MeasurementResult. The selection on which outcome to choose is based on internal RNG.

It would be useful to invert this to post-select and project on a set measurement, e.g. a signature

GeneralizedTableau.project(index:int, value:MeasurementResult) -> float

where the value is what is projected to, the return is the probability of being projected to that state, and the state is mutated to a new state given the MeasurementResult.

From what Claude and I can understand, this is exposed in Rust but not in Python. I can compute the probability via the expectation value of Z, but I cannot project. Instead, that requires GeneralizedTableau.fork and repeat until success, which admittedly feels a little hacky.

While non-physical, this would be super useful for modeling the action of stochastic programs as well as computing marginal bitstring probabilities $P(z) = |\langle \psi |z\rangle|^2$ through the chain rule. I have a use case that requires this marginal probability not to be mentioned in open source 😉.