Refactor HTML rewriter class to make it more open to change and expressive

[benoit74]

Fix #305

Idea: we should be able to define individual rewriting rules with functions and decorators, just like we do when defining endpoints in FastAPI. We have one decorator per kind of modification: drop attribute, rewrite attribute, rewrite data (and maybe others) and we implement one decorated function per kind of modifications expected.

Changes:

• new HTMLRewritingRules class with decorators drop_attribute (to decide to drop an HTML tag attribute), rewrite_attribute (to rewrite a given HTML tag attribute - name + value), rewrite_tag (to rewrite the whole HTML tag, potentially dropping it completely or changing tag name), rewrite_data (to rewrite HTML tag data)
• an instance of this class is wired in the html rewriting logic to properly process HTML code
• new tests are added to unit-test the new class behavior

See https://github.com/openzim/warc2zim/pull/353 for a show case of how this new code structure make it easier to:

• extend rewriting logic with support for new case
• test in isolation the extension, without having to deal with whole HTML tree

[benoit74]

@rgaudin this is far from being done, but I would like to have a first feedback on this approach

For now only the simple "drop attribute" case is implemented, but it gives a rough idea of what it will look like if I continue on this approach.

I'm a bit surprise I had to implement the call_func helper function myself, I'm pretty sure it is possible without this custom code and Python stdlib code, but I failed to find how to write it properly. Idea is that I want to pass whatever argument are defined in the decorated function but not other ones.

[benoit74]

Refactoring work is now completed, all HTML rewrite operations are isolated in standalone methods.