Add automaticLocalSeed in Math.Random and utilize in noise blocks

[beutlich]

There was some in and out development going on in #1957 when the noise blocks and functions were introduced in MSL v3.2.2 with the automaticLocalSeed left to do. This pull request fixes #1861.

@akloeckner @dzimmer @fvanderlinden FYI

[HansOlsson]

Two comments:

• Having a predictable sequence based on instance name is good, but as I understand we currently have the possibility to generate different results when re-running the same model - and I believe we should keep that possibility.
• Why is the hashing function spending time to split off the component-path? Unless the hash-function is really broken a fixed prefix doesn't cause any problems. If the goal is to generate the same result regardless of simulation model name I believe that will fail for some quoted identifiers.

[beutlich]

Having a predictable sequence based on instance name is good, but as I understand we currently have the possibility to generate different results when re-running the same model - and I believe we should keep that possibility.

I guess it is the other way round as it was an unintended side-effect that the automatically generated local seed somehow depended on the number/order of noise blocks in the model. This is now circumvented in such a way that this automatically generated local seed only depends on the component name, no matter how many noise blocks are found in the model

Why is the hashing function spending time to split off the component-path? Unless the hash-function is really broken a fixed prefix doesn't cause any problems. If the goal is to generate the same result regardless of simulation model name I believe that will fail for some quoted identifiers.

True. See also #1957 (comment). I'll fix automaticLocalSeed to use the plain path input as hashString input.

[HansOlsson]

I guess it is the other way round as it was an unintended side-effect that the autmatically generated local seed somehow depended on the number/order of noise blocks in the model. This is now circumvented in such a way that this autmatically generated local seed only depends on the component name, no matter how many noise blocks are found in the model

Ah, yes, I see - we have kept that possibility; just more predictably.

[HansOlsson]

Looks good now.