Simplify passing the arguments to OrderedDict in lexer

[plamut]

Token definitions in the custom parser for bigquery cell magic are passed as a list of (key, value) pairs, which results in extra two levels of indentation after formatting and additional noise (parentheses, quotes...).

Since we don't support Python 3.5 and below anymore, we can define the tokens as kwargs to OrderedDict. The kwargs ordered in Python 3.6+ is preserved (PEP 468).