add random behavior for size of sink request

[nuclearkatie]

• Three new optional parameters to vary the size of the sink's request
• random_size has default None, but also four two options that call the PRNG from the context:
  • "UniformReal", "NormalReal"
  • No longer implemented
    • "UniformInt", "NormalInt"
• When using one of the normal distributions, two additional parameters random_size_mean and random_size_stddev are used as fractions of the maximum request size (calculated at each timestep RequestAmt())

Checklist

• testing
  closes Design unit test for randomness in request size for Sink archetype #556

[gonuke]

If this is going into Cycamore, we might want to discuss what random sink behaviors make sense. I know you want a way to demo them all, but not sure any "Int" method makes sense?

[nuclearkatie]

After deciding if fractional or fixed μ and σ makes more sense, my next question is how (if?) to restrict the user's choice of μ and σ. I expect few use cases would want to set the mean higher than the maximum allowable value (amt, which is the min of throughput and remaining inventory space), but do we need to actually restrict the user from setting that if that's their intent?

On the other hand, I worry that a lack of restrictions will result in users accidentally setting a distribution with a negligible probability of picking an allowable value-- we don't want to have to sample 100,000 times to get one value

[nuclearkatie]

Cyclus version used for github actions needs to be updated with the PRNG (cyclus/cyclus#1599) for the build/test workflow to pass

[gonuke]

Can you rebase this against main to update the testing? In the meantime we ( @bennibbelink ) need to sort out which images are available for testing downstream (cyclus/cyclus#1628)

[nuclearkatie]

@gonuke this branch is already rebased against main, including the last update #549

[gonuke]

OK - I was thrown off by seeing failures of the CircleCI tests... but maybe those are lingering and not relevant.