fix: Ensure CSS support is checked more robustly

[mydea]

Noticed this while debugging tests in sentry-javascript. it seems jsdom has a weird support for CSSMediaRule, where it is defined but if you look at CSSMediaRule.prototype it is:

CSSRule {
        parentRule: null,
        parentStyleSheet: null,
        constructor: [Function: CSSMediaRule],
        type: 4
      }

I guess this could also be the case for other "weird" browsers or engines. I think it's safer to make sure everything we need is actually there at this point.

ref getsentry#42

[Juice10]

@mydea Thanks for submitting this PR, I had no idea rrweb was being used to record virtual pages with JSDOM, very interesting use case!
I added a small suggestion, if you go ahead and merge that suggestion, I'd be happy to approve this PR and get it merged.

[mydea]

@mydea Thanks for submitting this PR, I had no idea rrweb was being used to record virtual pages with JSDOM, very interesting use case! I added a small suggestion, if you go ahead and merge that suggestion, I'd be happy to approve this PR and get it merged.

Yeah, I only really found it due to tests using jsdom. But it showed that apparently some older CSS spec had this different, which may also happen in other browsers I guess - see: NV/CSSOM#105

I commited your suggestion 👍 thanks!

[mydea]

Oops sorry accidentally deleted this branch 😬 restored it and reopened the PR!

[YunFeng0817]

@Juice10 Is it OK to merge this one?

[mydea]

OK, actually we ran into a problem with this (see getsentry#47), so I updated this PR to ensure we do not access window in module scope.

Now, this is only checked on-demand. I also split this into to methods, one to check when we only want to access this, and one when we want to monkey patch this.

[Juice10]

@mydea thanks for the refactor and the bugfix, I’ve merged the PR