Worker cleanup

[annevk]

I've been looking at workers today.

• base URL should be response's url https://www.w3.org/Bugs/Public/show_bug.cgi?id=28835 My idea for fixing this is to make the environment settings object return global object's [[url]]. We mutate global object's [[url]] after fetching.
• SharedWorker needs to be keyed on input URL (not location) as otherwise redirects make for a confusing and racy model. Implementations also do not match the specification here and use input URL. We should probably make it more explicit what the SharedWorker map is and where it lives. (Unit of browsing contexts?)
• Need to investigate whether we want to keep URLMismatchError. Gecko gets away with not implementing it.
• Kill a worker and terminate a worker should xref "in parallel".
• Need to decide on a global object IDL strategy: https://www.w3.org/Bugs/Public/show_bug.cgi?id=28765#c3.
• Inline URLReadonlyUtils per discussion with @domenic.
• Can we forbid username/password in the URL still?
• SharedWorker name argument defaulting should move to IDL.
• Look into whether fragment identifiers are ignored when comparing URLs.
• Define 404 handling for importScripts() better: https://www.w3.org/Bugs/Public/show_bug.cgi?id=27276.
• Define muted errors flag for importScripts(): https://www.w3.org/Bugs/Public/show_bug.cgi?id=28961.

Anything else?

[annevk]

Fragment identifiers are included as part of the constructor url and as part of self.location. (As @Ms2ger said Chrome does indeed not consider redirects for self.location, Firefox does.)

[annevk]

@kinu, @horo-t, @sof, thoughts on removing URLMismatchError in Chromium?