Allow setting different seeds for different QPUs

[ghanem-nv]

Required prerequisites

• [X] Search the issue tracker to check if your feature has already been mentioned or rejected in other issues.

Describe the feature

Currently, the simulators on different QPUs are all initialized with the same random seed. By using the same seed, samples from different QPUs are statistically correlated. Depending on the specific use case, this correlation may or may not be an issue, but it is better to avoid it by using different seeds.

An extreme example of this correlation (but admittedly, not a practical use case of QPUs) is when the same circuit is sampled on different QPUs. In this case, the samples are identical (perfectly correlated). If one then uses the usual formula for estimating an error bar, it would underestimate it because it assumes samples are independent and does not account for the correlations.