Results: metadata storage

[ceholden]

We want to store our run metadata in our configuration files, but yatsm line is currently quite selective of what metadata is stored (e.g., is is hard coded what gets saved). Furthermore, the metadata attributes it saves are stored separate from one other as separate "sub-files" in the NumPy "npz" save file.

Switch to saving all metadata, except the version, into one dictionary:

> result.files
['record', 'version', 'metadata']
> type(results.metadata.item())
dict
> results.metadata.item()
{
'YATSM': {
    'design_matrix': '1 + x + harm(x, 1)'
}
'CCDCesque': {
    'fit': {
        'asdf': None
    }, 
    'init': {
        'consecutive': 5
    }
},
'dataset': {
        'input_file': 'asdf'
},
'phenology': {
        'fit': {
            'test': 5
        }, 
        'init': {
            'red': 2
        }
}

This proposal is backward incompatible when looking for "design_info" and "design_matrix" in yatsm.cli.map.find_result_attributes, so handle both past and future cases.

[ceholden]

Waiting to close: need to update TSTools for new metadata storage method.