Issue 1183 fast cif parsing

[ltalirz]

Fix #1183

This 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.

Example:

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

• add cif.py to linter (1st commit)
• add scan_type (2nd commit)
• add test to check that results are identical (2nd commit)

[ltalirz]

@giovannipizzi
In order to have the default value of scan_type properly set, I had to both extend _set_defauls
and make scan_type a property, which I find rather ugly (if I only provide _set_defaults, I have no control over the order in which the set_ methods are called and I get an error because set_scan_type is called too late).

Suggestions for improvement welcome.

[giovannipizzi]

Is there a way to check that it is actually using flex as a scan type?
Maybe checking the attribute, but maybe it is not even enough, in case the parsing happens before the scan_type is read? It depends on whether the parsing occurs at init or at _prepare_cif.

[ltalirz]

yes, it actually isn't using flex. I will put the default values in the constructor of CifData instead.

[ltalirz]

It's still not quite correct...
I will simply use the constructor as regular people do ;-)

[giovannipizzi]

I'm not sure I understand what happens when you don't use the property. Can't you use self.get_attr('scan_type') instead of self.scan_type?

[ltalirz]

Closing this in favor of a new PR including also #1186