Add dynamical as a priority/source

[williamhobbs]

Closes #68, #2, #56 (after it was re-opened), #87, and #89.

To-do list:

• run a few more tests (and one day add actual tests...)
• update documentation
• add some filters with helpful error messages for incompatible combinations of inputs
• consider condensing functions (e.g., merge get_solar_forecast and get_solar_forecast_fast into one with use_fastherbie as an input parameter, maybe also merge the ensemble function as well...) I decided not to do this, at least for now.
• update remaining interpolation to center of hour (HRRR is maybe the only one left?)
• shift 1min clear sky to the 30s mark? (1min clear sky calculations should be centered on the 30s mark #89)

[williamhobbs]

For filters/checks/warnings on input parameter combinations, combinations that don't work:

• decomp_model == 'erbs' and model in ['hrrr', 'cams'] -> warnings.warn(f'model={model} does not require decomposition, but you entered decomp_model="{decomp_model}". This will not do anything.)
• hrrr_hour_middle == False and model != 'hrrr' -> warnings.warn(f'You entered hrrr_hour_middle=False, which does not apply to the model you entered, "{model}". This will not do anything.)
• hrrr_coursen_window not None and model != 'hrrr' -> warnings.warn(f'You entered hrrr_coursen_window={hrrr_coursen_window}, which does not apply to the model you entered, "{model}". This will not do anything.)
• [model == 'cams' and priority not None...]
• [model not an ensemble and entered member not None]
• [model not 'cams' and (cams_area not None or cams_api_key not None)]

[williamhobbs]

[edit: this was fixed with 2c3adf1]

This bit of code is not working for non-dynamical sources:

hefty/src/hefty/solar.py

Lines 1746 to 1764 in 7613eba

	if not get_ens_temp and not get_ens_wind:
	_, _, _, search_str_2 = model_input_formatter(
	init_date, run_length, lead_time_to_start, model,
	resource_type, full_ens=False, get_ens_temp=False,
	get_ens_wind=False, member=0)
	# 2m temperature search string should be second
	search_str_cust = "|".join(search_str_2.split('|')[1:2])
	# use control member
	member_cust = 0
	df_rem = get_fcast_dataframe(
	latitude, longitude, init_date, fxx_range, model,
	search_str_cust, priority, product,
	fast, attempts, resource_type,
	member_cust)
	# merge
	df_merged = pd.merge(
	df_merged,
	df_rem.drop(columns=['lead_time', 'number']),
	on=['valid_time', 'point'])

get_fcast_dataframe is getting temperatures for all members, not just the control. An example of this:

member_cust = 0
resource_type = 'solar'
latitude = [32, 30]
longitude = [-89, -90]
init_date = pd.Timestamp('2026-09-10 00:00:00')
fxx_range = range(9, 16, 3)
model = 'ifs_ens'
search_str_cust = ':2t'
priority = 'azure'
product = 'enfo'
fast=True
attempts=2

df_rem = get_fcast_dataframe(
                latitude, longitude, init_date, fxx_range, model,
                search_str_cust, priority, product,
                fast, attempts, resource_type,
                member_cust)

df_rem

returns a dataframe with 300 rows, but it should be 6.

To get just the control from IFS ens, options are:

• before 2026-05-12, use a search string like ':2t:sfc:g'
• after 2026-05-12, switch product to 'oper' and search string '':2t'' (or ':2t:sfc:g')

I think the solution is to use the product that is returned from model_input_formatter.

[williamhobbs]

Informal tests included:

• run through all the example notebooks (no errors, results look consistent) [edit: apparently I didn't wait for everything to finish; there was an error with cams, but I fixed it with the next commit...]
• check the fix for the last comment with something like:

resource = get_solar_forecast_ensemble(
    latitude=[32, 30], longitude=[-89, 90],
    init_date='2026-09-10', run_length=6,
    lead_time_to_start=12, model='ifs_ens',
    attempts=2, priority='azure')

• running some internal applications that use hefty