Refactor the parser in a single NetscapeBookmarkParser class

[virtualtam]

Modifications:

• update Composer autoload
• encapsulate all methods in a single class
• update tests
• update README and example

[virtualtam]

Ping @ArthurHoaro & @nodiscc , as this parser should be mature enough to be used within Shaarli ;-)

[ArthurHoaro]

Great! :+1: I've asked a few question about how it handles tags, but almost ready.

Also we'll have to be careful about empty URL and default timestamp with Shaarli integration.

[virtualtam]

Thanks for the review!

Regarding Shaarli-specific formatting:

• there are some notes in the test data (the "URL" only is the permalink part)
• I think there is a hack on the timestamps in Shaarli's index.php, that allows importing several links with the same timestamp

I think it'd also be worth simplifying the parser's cleanup steps, all those regexes feel a bit superfluous (=> separate PR)

[ArthurHoaro]

Permalinks are fine, we have to support it, to move a Shaarli to another for example. I was talking about this one.

[virtualtam]

I was talking about this one.

Hmm, I wouldn't worry too much about "linkless links", the only way one would end with this is by manually editing a bookmark dump and deleting the href attribute, which is kinda silly...

[virtualtam]

To sum up:

• 15 to improve packaging metadata

• 16 to add test cases & improve robustness a little bit

• 17 to simplify data cleanup (and maybe remove dead/useless code)