Implement new RNG scheme

[jared321]

Tentative PR Self-review

• Review all changes made here
• Confirm test script fails with invalid bit generator name
• Confirm that new MH benchmark is acceptable
  • New benchmark is deterministic as expected
• Confirm sampler test script passing with no changes except for one altered test.
• Confirm that ball drop notebook is now deterministic when run locally
  • Altered seed to see difference pop up. Reset seed to see no meaningful differences
• Review new/updated documentation as rendered by RTD for this PR
• Review contents of both notebooks as rendered locally
• Confirm all actions passing

[jared321]

@mosesyhc Can you please propagate the use of the new global surmise RNG throughout the code base. Please see the changes made so far to see what I have already done (nbdime, which is installed by the tox book task, might be useful).

• The pytests are failing because they aren't setting an RNG in surmise
• Need to convert all random variable draws (aside from in the samplers) to use this global RNG
• Iterate on the RNG content is User Guide --- example code needs to be redone
• Iterate on the RNG content in the Dev Guide

[mosesyhc]

A Jupyterbook page is added to illustrate the use of surmise RNG and reproducible results. Guides and README are modified to include a link to the Jupyterbook (minimally here, the JB reference should be made explicit and prominent in the release.)

@jared321 ready for your review. (realizing that this will be a self-review because you initiated the PR)

[mosesyhc]

Reviewed via

• the running of Jupyterbook
• manually clear RNG and confirm failure
• set same RNG seeds to confirm reproducibility
• set different RNG seeds to confirm that different samples are drawn