Tighten number lexing

* Tokenize strings of the form "7E+X", "2.3E " etc as a number
  (integer, real) followed by a new token starting with 'E'.
  Note that this makes the string "1.2E3E4" no longer cause a
  lexer error status.
* Add tests for corresponding numeric disambiguation in
  `text_lexer.cpp`.
* Add tests for stoichiometric expressions that ensure that only
  otherwise ambiguous phrases such as "7E+2F" fail to parse
  ("7E+F" should parse correctly as a stoich expression.)
* Add missing comma in list of 'good' expressions in
  the `Parser.parse_line_expressions` test.