fix: Only use HTML rules if mimeType matches

[karfau]

In the living specs for parsing XML and HTML, that this library is trying to implement,
there is a distinction between the different types of documents being parsed:
There are quite some rules that are different for parsing, constructing and serializing XML vs HTML documents.

So far xmldom was always "detecting" whether "the HTML rules should be applied" by looking at the current namespace. So from the first time an the HTML default namespace (http://www.w3.org/1999/xhtml) was found, every node was treated as being part of an HTML document. This misconception is the root cause for quite some reported bugs.

BREAKING CHANGE: HTML rules are no longer applied just because of the namespace, but require the mimeType argument passed to DOMParser.parseFromString(source, mimeType) to match 'text/html'. Doing so implies all rules for handling casing for tag and attribute names when parsing, creation of nodes and searching nodes.

BREAKING CHANGE: Correct the return type of DOMParser.parseFromString to Document | undefined. In case of parsing errors it was always possible that "the returned Document" has not been created. In case you are using Typescript you now need to handle those cases.

BREAKING CHANGE: The instance property DOMParser.options is no longer available, instead use the individual readonly property per option (assign, domHandler, errorHandler, normalizeLineEndings, locator, xmlns). Those also provides the default value if the option was not passed. The 'locator' option is now just a boolean (default remains true).

BREAKING CHANGE: The following methods no longer allow a (non spec compliant) boolean argument to toggle "HTML rules":

• XMLSerializer.serializeToString
• Node.toString
• Document.toString

The following interfaces have been implemented:
DOMImplementation now implements all methods defined in the DOM spec, but not all of the behavior is implemented (see docstring):

• createDocument creates an "XML Document" (prototype: Document, property type is 'xml')
• createHTMLDocument creates an "HTML Document" (type/prototype: Document, property type is 'html').
  • when no argument is passed or the first argument is a string, the basic nodes for an HTML structure are created, as specified
  • when the first argument is false no child nodes are created

Document now has two new readonly properties as specified in the DOM spec:

• contentType which is the mime-type that was used to create the document
• type which is either the string literal 'xml' or 'html'

MIME_TYPE (/lib/conventions.js):

• hasDefaultHTMLNamespace test if the provided string is one of the miem types that implies the default HTML namespace: text/html or application/xhtml+xml

[SmartLayer]

i'm very eager to test this when merged and a pre-release becomes available

[karfau]

@weiwu-zhang Thank you for the feedback.
I still have some WIP locally that I need to cover with tests, before this can land.
Sadly I didn't have time for this subject in the last week, but I will be back on it soon.
Of course no time line promises.