Use same RNG across chains in transformed data block

[bob-carpenter]

Summary:

Use the base RNG (with no offset for chain) for each chain in the transformed data block.

Use no skipping for transformed data, then skip plus one for each chain (so minimum skip is one unit).

Description:

Couldn't think of use case for having different RNGs across chains as we have now as it will not allow convergence monitoring or inference iif the data's different in different chains.

This is to be staged after the services refactor so that it can be done in this repo rather than in each interface.

Test

Make sure multiple chains with fixed see dand chain ids starting at zero all get the same random transformed data (can save it in generated quantities block for testing).

Current Output:

Different transformed data in each chain.

Expected Output:

Same transformed data.

Additional Information:

Provide any additional information here.

Current Version:

v2.11.0

[bob-carpenter]

If someone can help me find where the model is instantiated with an RNG from the services, I can take care of this one.

[bob-carpenter]

OK, this isn't an issue any more, but there's a new one: #2241