It's not that the test does not work, but it's probably a bad idea to
spam the server with our requests. One should configure a test instance
with low frequency run to test this functionnality

In the test environnement, relative import of Spam return
"pastealacon.browser.Spam" object, that we compare with
"module.pastealacon.browser.Spam". It cannot work.

I'm not sure of the root issue of this bug, bug fix it to pass the test.

This reverts commit 15859f32.

weboob.tools.browser -> weboob.deprecated.browser
weboob.tools.parsers -> weboob.deprecated.browser.parsers
weboob.tools.mech -> weboob.deprecated.mech
weboob.browser2 -> weboob.browser
weboob.core.exceptions -> weboob.exceptions

Also, the new tree for browser2 is:

weboob.browser: import weboob.browser.browsers.* and weboob.browser.url.*
weboob.browser.browsers: all browsers (including PagesBrowser and LoginBrowser)
weboob.browser.url: the URL class
weboob.browser.profiles: all Profile classes
weboob.browser.sessions: WeboobSession and FuturesSession
weboob.browser.cookies: that's a cookies thing
weboob.browser.pages: all Page and derivated classes, and Form class
weboob.browser.exceptions: specific browser exceptions
weboob.browser.elements: AbstractElement classes, and 'method' decorator
weboob.browser.filters.*: all filters

One of the goal is to not import all modules needed by filters by
loading the page file.

In the same goal, move the import of parsers in the class definition.

This also transforms the fill_ function into both a filling and a getting
function. Magic!

And autodetect it on forms. There is no other way to know what is the
expected encoding.

It works, but there are a lot of things that could be improved.