Add cloud utilities for multirows, datalinks

[zoghbi-a]

With the goal of breaking the other PR #18 into smaller pieces. The code here processes the cloud information, and it uses the download functions in PR #21 (so that should be approved first).

The main parts:

• Function to look for and process the cloud information: _process_json_column, _process_cloud_datalinks and _process_ucd_column. These return a dictionary elements: e.g. {'url':..} for prem or {'uri':.., 'bucket_name':.., 'key': ...} for aws that can be passed to the relevant download functions.
• The user facing function get_data_product2. This is similar to what we have now but handles datalinks correctly, and multiple Record's (QueryResult) and Row's (Table). Once review is done, the name should be get_data_product.
• get_data_product takes a product (pyvo Record or astropy Row/Table) and a provider (prem, aws etc) and a returns a ProviderHandler instance (a list that has a download method).
• get_data_product2 takes also another option mode if the user wants to select json, datalink (others can be added).
• ProviderHandler has also a get_links method that can be used to figure out the url/uri in case the user want to use the cloud cutout for example.
• All the extra code is compared to Add download utilities #21 is in _fornax.py so not to interfere with fornax.py, but once review is done, the first should replace the latter.

An example use case is:

# Query data provider 
chandra_url = 'https://heasarc.gsfc.nasa.gov/xamin_aws/vo/sia?table=chanmaster'
chandra_service = pyvo.dal.sia.SIAService(chandra_url)
query_result = chandra_service.search(pos=(pos.ra.deg, pos.dec.deg), size=10*u.arcsec)

# prem data
chandra_prem = fornax.get_data_product2(chandra_result[0], 'prem')
chandra_file = chandra_prem.download()

# aws data
chandra_aws = fornax.get_data_product2(query_result[0], 'aws')
chandra_file = chandra_aws.download()