Handling for module-level dimension variables for interstitial variables

[peverwhee]

Updates to allow for dimension variables that are set in the register phase to be used to allocate interstitial variables.

Open to feedback, of course! There's no explicit handling right now to ensure that the variable is set in the register phase, but my feeling is that you'll get a run-time error if you try to use an uninitialized variable in an allocate statement.

User interface changes?: No

Fixes:
closes #681

Testing: Added variables/logic in capgen test to confirm the dimensions are initialized and used properly

[climbfuji]

Looks reasonable to me. And I agree that if the dimension isn't set, we'll get a runtime error. Do you think the auto-generated code that is added with --debug will catch it in the caps?

[climbfuji]

Can you update the docstring please?

[peverwhee]

updated!

[peverwhee]

@climbfuji good question!

I tested this by moving the setting of the dim_inter variable in the capgen test to the init phase instead of the register phase (which means the module-level interstitial variable will be allocated before its dimension is set later in the init phase). Can confirm that it does run up against the debug checks and produces an error.

physics2        : In group temp_suite_finalize before  temp_adjust_finalize: for array interstitial_var, expected size        3 but got        0

[climbfuji]

@climbfuji good question!

I tested this by moving the setting of the dim_inter variable in the capgen test to the init phase instead of the register phase (which means the module-level interstitial variable will be allocated before its dimension is set later in the init phase). Can confirm that it does run up against the debug checks and produces an error.

physics2        : In group temp_suite_finalize before  temp_adjust_finalize: for array interstitial_var, expected size        3 but got        0

Great, thanks for testing!

[gold2718]

I think this is okay.

[peverwhee]

@mkavulich I believe this can be merged? unless I missed a discussion about this at the last framework meeting