Add 'k' extension to SMARTS to support ringsize queries

[greglandrum]

The r primitive in SMARTS is for minimum SSSR ring size, so for this molecule:

we get results like this:

In [5]: m = Chem.MolFromSmiles('C1CC2N1CCCCCC2')

In [6]: m.GetSubstructMatches(Chem.MolFromSmarts('[r8]'))
Out[6]: ((4,), (5,), (6,), (7,), (8,), (9,))

i.e. atoms 2 and 3 do not match the r8 search; they would match an r4 search:

In [3]: m.GetSubstructMatches(Chem.MolFromSmarts('[r4]'))
Out[3]: ((0,), (1,), (2,), (3,))

This PR adds the new primitive k that just uses SSSR ring size:

In [7]: m.GetSubstructMatches(Chem.MolFromSmarts('[k8]'))
Out[7]: ((2,), (3,), (4,), (5,), (6,), (7,), (8,), (9,))

This is part of fixing #9143, but I think it will be generally useful.

[ricrogz]

LGTM. Just a small suggestion to use an enum instead of a string.

[ricrogz]

Given the limited number of choices here, and that the prefix is never shown to the user, would it make sense to use an enum class for this, like:

enum class : char  { EQUAL, LESS, GREATER, RANGE };

[greglandrum]

LGTM. Just a small suggestion to use an enum instead of a string.

Makes sense. The suggested changes are in