Basic Ability to Integrate with TSM

[sjordan29]

I propose merging this initial approach for integrating TSM and Clearwater-Riverine. This PR includes the following:

• Addition of an initialize() method consistent with BMI. Ultimately, this can get linked with the ClearwaterRiverine init, but for now I am keeping them separate to allow for easier testing since initialization can be slow with many timesteps. We can revisit this when Speed up calculation of the SUM_OF_COEFFICIENTS_TO_DIFFUSION_TERM #49 is closed.
• Addition of an update() method consistent with BMI. The previous code only allowed users to run the entire model at once. This new method allows users to increment by a single timestep.
• Modifications to the plot() method and its dependencies to work with the new incremental timestep approach.
• Preserves simulate_wq() so that old demos do not break.

These additions allow clearwater riverine to run in a loop with clearwater modules for TSM, as demonstrated in examples/dev_sandbox/43_Realistic_TSM_Scenario.ipynb. I propose that this close #43 and that we open more specific issues as needed as we polish that demo.

[aufdenkampe]

@sjordan29, this PR looks great. It demonstrates the capability to couple with TSM and profiles the performance of that coupling test case, setting us up nicely for the next round of work on:

• Speed up calculation of the SUM_OF_COEFFICIENTS_TO_DIFFUSION_TERM #49
• Improve timestep performance #52
• Improve timestep performance further: numba or not? ClearWater-modules#57