API to invalidate BFCache'd webpage with outdated content added by the extension

[tophf]

Problem:

1. Use an extension that modifies DOM of webpages e.g. applies themes (dark mode), adds DOM stuff like toolbars, etc.
2. Navigate to another page, so the modified one goes into bfcache
3. Change an option in the extension that substantially alters the way it affects DOM
4. Go back to the bfcache'd page and observe the completely wrong content/appearance.

If we try to use a listener on pageshow event and retrieve the configuration, there'll be a noticeable delay before the up-to-date configuration is applied, which is very jarring when the visual changes are substantial.

Possible solution:

chrome.tabs.invalidateBFCache(tabId); // all cached documents
chrome.tabs.invalidateBFCache(tabId, {documentId: '...........'}); // just one

Alternatives:

chrome.tabs.update(tabId, {removeBFCache: true});
chrome.tabs.update(tabId, {removeBFCache: documentId1});
chrome.tabs.update(tabId, {removeBFCache: [documentId1, document2]});

[tophf]

A potentially [much] simpler and cleaner solution might be to allow chrome.scripting.executeScript on a bfcached documentId so it will run when the document is restored:

chrome.scripting.executeScript({
  target: {tabId, documentIds: ['.........']},
  func: config => {/* apply config to the page */},
  args: [config],
});

Right now it ignores documentIds and just runs in the current document of the tab.

[dotproto]

Note from today's meeting WECG meeting: using extension messaging to post a message to a BFCached document used to evict a document, but doesn't any more.

[Rob--W]

The general consensus is to not introduce APIs to enable extensions to blanket evict bfcache'd contexts (meeting notes pending review at #750).

Instead, the idea is that the pageshow event is thought to be the right API, and that the problem of latency could be addressed by introducing an API to update values that can be retrieved synchronously from the content script. That topic has been discussed many times before, and #536 is one of the issues related to it.

[tophf]

Indeed, config registration would help, I totally forgot about it. I vaguely remember now it was about to be implemented 5+ years ago in Chrome.