Additional arguments to `create_time_discretization` methods

In GitLab by @Huite on Jun 21, 2023, 15:06

The time discretization objects store duration of stress periods, the number of timesteps (generally 1.0) and a multiplication factor (for expontential growth).

We require the create_time_discretization method on the model or simulation, as it checks all packages and derives stress period definition from there. However, it's not uncommon to have one or a few steady-state periods in a simulation (especially at the start). This currently requires something like this:

model.create_time_discretization(["2000-01-01", "2010-01-01"])
model["time_discretization"]["transient"][0] = False

I would argue that something like this is nicer:

model.create_time_discretization(transient={"2000-01-01": False})

And similarly for the other settings:

model.create_time_discretization(transient={"2000-01-01": False}, n_timesteps={"2001-01-01": 5})

It's arguable whether the transient argument should be a dict, since its values are always boolean. It could also be a list of steady-state dates, but having dicts like these are consistent across the arguments.

The default values can remain the same: almost all stress periods are transient=True, n_timesteps is nearly always 1, and the multiplication is nearly always 1.0.
In the worst case, a user can still access the entire array in the "old way":

model["time_discretization"]["transient"][:] = False

Start by locating the model and simulation create_time_discretization methods and inspect how they derive the transient, n_timesteps, and multiplication settings. Define done as supporting the proposed date-keyed overrides while preserving the existing defaults and direct array access; the issue does not name specific files or tests.