Extend protocols to JS built ins using CLJS "type symbols", fix extending protocols to default & Object

[lilactown]

Please answer the following questions and leave the below in as part of your PR.

• I have read the developer documentation.
• This PR corresponds to an issue with a clear problem statement.
• This PR contains a test to prevent against future regressions

---

Fixes #781, #782 and #783

[lilactown]

even in the impl before this PR, this would AFAICT never happen because it wasn't possible to extend a protocol to either 'default or 'object

[borkdude]

@lilactown I tried this with this PR checked out locally:

cljs.user=> (sci/eval-string "(defprotocol IFoo) (extend-type string IFoo) (satisfies? IFoo \"foo\")" {:classes {'js goog/global :allow :all}})
false

[lilactown]

@borkdude I think that's due to another bug - that satisfies? fails to find protocols that don't have methods. For instance, this works:

cljs.user=> (sci/eval-string "(defprotocol IFoo (foo [_])) (extend-type string IFoo (foo [_])) (satisfies? IFoo \"foo\") {:classes {'js goog/global :allow :all}})

And just to show that this isn't a regression, testing js/String using my originally installed version of nbb

Welcome to nbb v0.6.129!
user=> (defprotocol IFoo)
IFoo
user=> (extend-type js/String IFoo)
{:methods #{}, :name user/IFoo, :ns #object[Eq user], :sigs {}, :var #'user/IFoo, :satisfies #{function String() { [native code] }}}
user=> (satisfies? IFoo "foo")
false

[lilactown]

Created #785 to track the bug with satisfies? + no methods

[borkdude]

Thanks!