CifData: fast scanning and lazy parsing

[ltalirz]

Fix #1183
Fix #1186

Fast scanning

Adds a scan_type attribute to CifData.
Using scan_type='flex' uses C-routines of PyCifRW and is several times faster than the python-based implementation for large CIF files.
The default behavior is unchanged.

Example:

from aiida.orm.data.cif import CifData
a = CifData('/path/to/file.cif', scan_type='flex')

Lazy parsing

Adds a parse_policy attribute to CifData.
Using parse_policy='lazy' delays parsing of CIF files until the values are actually needed.
The default behavior is unchanged, except for the fact that the CIF object is now kept in memory after regular parsing (see #1186) (it will, however, not be written to the DB).

Example:

from aiida.orm.data.cif import CifData
a = CifData('/path/to/file.cif', parse_policy='lazy')

Note: 1st commit contains changes from formatter & linter

[ltalirz]

One change is that the calls to get_formula and get_spacegroup_numbers have moved from set_file into the constructor.
This means, the behavior of a = CifData('/path/to/file.cif' remains the same, but the behavior of
a.set_file('/path/to/file.cif') changes.

The reason for this change is that otherwise one needs to start fighting against the "cleverness" of AiiDA, which calls all the set_... functions in the constructor of the base class (e.g. one would need to make sure that set_parse_policy to be set before set_file).

I'll add something to set_file in order to prevent inconsistencies.

[ltalirz]

Closing this PR until convergence reached.

[ltalirz]

@giovannipizzi Feedback appreciated :-) You are also welcome to assign someone else, if you don't have the time.

[ltalirz]

@nmounet If you are not available, please unassign yourself and I will ask someone else.

[ltalirz]

@giovannipizzi It seems the Nicolas currently has no time.
Should we go ahead with this? I'm happy to add a bit of documentation here as well, if we agree on the implementation.

[giovannipizzi]

Do you know why tests are failing? (Probably when you'll fix them I have to approve again)

[ltalirz]

Yes... I removed the main_file_name argument of _prepare_cif because it is not used (the function just returns the cif string).
Now I see that there are many similar functions like this... I guess removing it is the right thing to do, but it might break other code...