New option for : lexer keyword=true

This option is complementary to the option emitOnMatch, but I think more complicated to implement.

If I want to parse
http://www.lemonde.frhttp://www.bbc.co.ukhttp://www.httpsite.com => SITE (www.lemonde.fr) SITE ( www.bbc.co.uk ) SITE (www.httpsite.com)
with rules

```
HTTP : ( option keyword = true : 'http://' ) -> skip;
SITE : SITECHAR+;
fragment SITECHAR : [a-zA-Z0-9\.];
```

`http://` would be considered as a keyword and would act as a delimiter. If matched then the lexer matches the first rule with characters before the 'h' of `http://`.
It means the sequence `http://` is tried for each new character read by the lexer, and the rule HTTP is never discarded.

No files or tests are named. Start by reviewing the lexer handling for the existing emitOnMatch option and how lexer rules are retained or discarded; done means the keyword rule is tried at each new character and delimits the preceding SITE text as shown in the example.

Written by the indexing model from the issue text.