Add iotools function to write SAM-formatted CSV files

[williamhobbs]

Is your feature request related to a problem? Please describe.
Manual formatting is required before writing weather data into a CSV file that NREL SAM can read.

Describe the solution you'd like
An iotools function, e.g., pvlib.iotools.write_sam(), that writes data to a CSV file in SAM's format in a standardized way would be nice.

Describe alternatives you've considered
Manually re-formatting data, adding relevant headers, and writing a CSV. An example of code I've used is below:

## Write SAM Weather file - SURFRAD Goodwin Creek 2019
# note: start with dataframes for data and metadata created with pvlib.iotools.read_surfrad, e.g., `(surfrad_data, surfrad_metadata) = pvlib.iotools.read_surfrad(...)`

sam_data = surfrad_data

# Make sure file starts at midnight local time. Here I use roll (circular shift) as a quick fix
sam_data = sam_data.reindex(index=np.roll(surfrad_data.index,-6*60))    

# metadata
source='SURFRAD Goodwin Creek 2019'
latitude=surfrad_metadata.loc['latitude',0]
longitude=surfrad_metadata.loc['longitude',0]
elevation=surfrad_metadata.loc['elevation',0]
timezone_offset = -6 # standard time in hours ahead of GMT, e.g., -7 for Denver

filename = source + '.csv'

# make a header        
header = '\n'.join(
        [ 'Source,Latitude,Longitude,Time Zone,Elevation',
        source + ',' + str(latitude) + ',' + str(longitude) + ',' + str(timezone_offset) + ',' + str(elevation),
        'year,month,day,hour,minute,ghi,dni,dhi,temperature,wspd',
        ]) + '\n'

# write 1st part of header
with open(filename, 'w') as fp:
    fp.write(header)

# write rest of header with SAM-friendly names and write weather data
columns = ['year','month','day','hour','minute','ghi','dni','dhi','temp_air','wind_speed']
sam_data.to_csv(filename, columns = columns, mode='a', index=False, header=False)

Additional context
There is an open issue in https://github.com/NREL/SAM/, NREL/SAM#866, to update SAM to allow weather variable names used by pvlib. That would simplify things.

There may need to be additional consideration given to weather datasets that:

• include less than one year, which currently will not work in SAM
• include a full year, but don't start January 1 at midnight. Automatically "rolling" could fix this
• include multiple years. The user would either have to pick the year to be written, or multiple weather files would need to be written, each containing one year (preferred).