Running drivers as `um2nc` subcommands

[blimlim]

Model drivers are currently using seperate entry points to the main um2nc command:

um2nc-standalone/pyproject.toml

Lines 29 to 31 in d7e4f4f

	[project.scripts]
	esm1p5_convert_nc = "um2nc.drivers.conversion_driver_esm1p5:main"
	um2nc = "um2nc.um2netcdf:main"

Following discussion with @atteggiani, we should look into running the drivers instead as subcommands of the main um2nc. For example, running the ESM1.5 conversion driver could look like um2nc esm1p5 <args>

This could simplify the interface, make it easier to pass um2nc processing arguments directly to the drivers rather than hard coding them, and allow for um2nc to be released as a seperate binary. This last point would help us decouple um2nc and payu releases, and allow us to avoid singularity containers and the associated transient errors.

[atteggiani]

Rather than having multiple subcommands for each driver, I suggest having a fixed subcommand (that accepts inputs) to determine if any specific driver logic needs to be run (instead of a "simple conversion"). The input driver can determine which driver script to use.
For example, the command could be run-driver, and it could be used as:
um2nc run-driver esm1p5 <esm1p5_driver_args> or um2nc run-driver esm1p6 <esm1p6_driver_args>.

[atteggiani]

With argparse subparsers we can manage different subcommands and also set required/optional arguments that can be:

• shared across the entire CLI ("simple conversion" and all "driver conversions" will have them)
• shared across the drivers ("simple conversion" does NOT allow them but all "driver conversions" will have them)
• only allowed for one specific driver (only the specific driver will allow that argument)