…ed tests. I should clarify that my dream was explaining that instead of force softening, using the time dialation scheme (dialating the advective term) has a similar property of constraining the time step due to advective restrictions without allowing extra leakage. Insead of force softening, which allows extra stuff to get through the throat, we dialate the time in this region to just made evolution there slower. It preserves the energy contours of the problem while allowing a larger time step. Yes, I did give this detailed description to Ammar in my dream including figures and diagrams.

- Added gkyl_array_min and gkyl_array_min_range functions for element-wise minimum operations.
- Implemented CUDA kernels for minimum operations on arrays.
- Updated tests to include new minimum functionality.
- Removed unused previous step CFL rate logic from species dynamics.

…erations and dg_bin ops. I made a special basis for this operation. Added in the ability for either users to specify the minimum dt, or for the code to minimize dt with the time dialation method based on omega_H. The user can always adjust the omega_H frequency at the input file level by specifying something like cfl_frac for the omegaH mode.

…ate into binary operations; update collisionless app for CFL scaling

… think this means we need to combine this into one big method because we have to allocate everything at initilization. Or at least the initilization needs to initilize everything. Tested the poa regression test and it still works. On to simulations!

…te related calls in tests and applications. Add and verify GPU unit tests for this function

…eshold to determine the minimum dt timestep restriction

…caling function

…ng to mask CFL. There were some GPU side issues with the array_reduce operation, but it is resolved. Remove all boolean operations for the arrays. Add a skip_cell_advance_inverse operation because using array arithmatic gives FLOP arithmatic errors.

…ask updater instead of the skip cell updater. It's cleaner now, plus I added an essential feature to the array mask to mask things that are some fraction of the maximum f value. I need to test this implementation more thoroughly, but it should be working now. This is just a scripting change in the app, so it should be fine. I updated the reset method to have a more robust decision process of releasing and re-initializing the array_mask so that it can change between the different thresholds and update the threshold

…ns for improved readability and consistency

…ier.I tested some regression tests and they are valgrind clean. First results look reasonable, but I can't really tell if it's working correctly as intended. The logic should all be inside the app

…s for each version of time dilation. This is because we only want the allreduce to happen for the local mask. There were also if statements in the advance method, which is a red flag. Format with uncrustify. Must test on GPU and CPU

…he time dilation

…ctor (so this is stored inside gk_species). This allows us to dilate electrons relative to ions to have much larger time steps and reduce the scale separation between them.

…device memory. It was doing the incorrect operation, bu not failing becaause it wasn't a computationally wrong thing, I just meant for it to do something else.

…d functions

- Introduced a new fdot multiplier type to set dt based on another species.
- Added functions to compute global maximum and clamp CFL rate by omega max.
- Refactored existing functions better clarity and maintainability.

…ypes and update related structures and functions