Add support for zero-order bond

[merkys]

Clark, 2011 argues for introduction of zero-order bonds to ease the representation of non-organic compounds. This suggestion was directed at MDL format, but I believe it is also applicable to SMILES. Currently the representation of metal coordination in SMILES is ambiguous: one may represent it using single bond or not represent interaction at all, losing an important piece of information.

As for the symbol, I think ! could be easy to remember as it resembles both a dashed line (used in Clark, (2011)) and asymmetry due to donor-acceptor interactions.