Merge: Loose Tokens

Another step to simplify the clients of AST: loose tokens.

Some tokens, correctly lexed by the lexer, are discarded by the parser because of (thanks to?) sablecc transformations. Mainly comments and new-lines (EOL) are lost but some other tokens might be also discarded (currently, commas and dots are still lost).

Previously, the full sequence of tokens can still be accessed trough `next_token` and `prev_token` but this mean that tools like nitprettty and nitlight have to maintain two cursors, one on the AST and one on this linked list of tokens and try to keep these two cursors synchronized.

This PR names the concept of *loose* tokens that are lexed tokens but absent from the AST, and attach to each (non-loose) token in the AST a list of loose tokens that precede it (`prev_looses`) and that follow it (`next_looses`).
This way, a visit on the AST can be used to also easily process all the tokens from the source.

Future PRs will try to use this new thing to improve/simplify/whatever nitpretty and nitlight.

Pull-Request: #1346
Reviewed-by: Alexandre Terrasa <alexandre@moz-code.org>
Reviewed-by: Lucas Bajolet <r4pass@hotmail.com>

diff --cc src/parser/parser_nodes.nit
Simple merge