Speedup for rolling functionality, support for uneven/latlon grids, testing/CI, and small changes

Speed-up from numpy.roll numpy.roll is relatively slow since it makes copies of the original array, so the rolling (or padding) functionality is now handled iteratively with indexing to shift the array with or without replacement, depending on the boundary conditions. The example below is the case for periodic boundary conditions.
```
for i in range(10): 
    xroll[:i] = x[-i:]
    xroll[i:] = x[:-i]
    adv_E_roll[:, :i] = adv_E[:, -i:]
    adv_E_roll[:, i:] = adv_E[:, :-i]
```
Support for more grid types (rectangular, uneven spacing, latitude-longitude)
- This support is primarily handled through function arguments.
- Rectangular grids will result in binned structure functions and separation distances for the zonal direction only
  - Number of bins can be set as a function argument
- In the future we could rebase to a more class-based structure.
Set up continuous integration with GitHub Actions to test and check formatting
- Using pytest for testing modules. Currently only testing calculate_velocity_advection with very simple test cases.
- Using Ruff as a linter with black as the formatting standard.
Miscellaneous updates
- Dictionary data output
- Gradient calculation updates throughout advection modules and structure function modules

Rectangular grid support for the meridional direction
More testing for the latlon and uneven grids -- still somewhat a work in progress
Work on getting the linter to pass
Add additional pytest tests for other modules.

cassidymwagner / fluidsf