Add script sanitizer to convert the most common scripts to AMP

[westonruter]

WordPress users very commonly add script tags to Custom HTML widgets and directly to theme templates to show ads, add analytics, etc. Assuming that the vast majority of these scripts are very templated (e.g. including via Google Tag Manager), then it should be possible to reliably identify the each script and extract the relevant information to convert to AMP.

See also #1031 which could serve as a way to build up a database of scripts for prioritization.

[westonruter]

As part of the script sanitizer this we could potentially look at de-duplicating validation errors that are currently coming when you enqueue some JS. For example, when enqueueing twentyseventeen-navigation there are two validation errors reported:

• enqueued_script
• invalid_element

The first is coming from when the script is first enqueued and the second comes from the whitelist sanitizer when the script is actually removed. This is due to the fact that the enqueued_script validation error is triggered prior to sanitization happening in the first place: it gets triggered during output buffering. When coming across an invalid_element error we could potentially look at AMP_Validation_Utils::$enqueued_script_sources to see if it exists there (by comparing the src with the enqueued script src URLs) and then skip reporting the validation error. But in this case should it also apply to any scripts that are printed due to being dependencies of the originally enqueued script?

[westonruter]

This is being added in 811d133#diff-7d234989187d87a372e2aa6aa1299a71R672 via #1093.

[westonruter]

It will be important to also make sure any noscript elements are removed if they have corresponding scripts that are recognized and converted to AMP.

[westonruter]

@amedina Here's another bit of code that should be moved into the new script sanitizer:

https://github.com/Automattic/amp-wp/blob/a68c93fd62df93c6c79cf99d4917bb912e11b684/includes/class-amp-theme-support.php#L1064-L1069

[westonruter]

Adding support for noscript contents replacing adjacent script elements could be a better first step for this (see #1213).

[westonruter]

Closing per resolution in #1127 (comment)