which is a regression compared to 0.8.x, most likely introduced as part
of #498

- add check to `parseDoctypeCommentOrCData`
- drop redundant and broken `Element.appendChild` implementation
- `hasInsertableNodeType` now checks for `CharacterData` nodes instead
of only text nodes
- align ParseError and DOMException in how they are extending Error
- wrap `DOMException`s in `ParseError` in sax parser
- move custom errors to own module
  - and allow current and modern constructor arguments for DOMException

Per XML spec §2.6, PI data extends to immediately before '?>'.
Trailing whitespace inside the PI boundary is content, not separator.
Remove \s* before $ in parseInstruction regex so it is no longer stripped.

Addresses the trailing-whitespace sub-issue from xmldom#42, backporting xmldom#498
behaviour to the 0.8.x line.

**Target branch**: `release-0.8.x`

### What

Remove `\s*` from the `parseInstruction` regex in `lib/sax.js` so that
trailing whitespace inside a processing instruction is preserved instead
of silently stripped.

### Why

Per [XML spec §2.6](https://www.w3.org/TR/xml/#sec-pi), PI data is
everything between the mandatory separator whitespace after the target
and the closing `?>`. Trailing whitespace inside the PI boundary is
content — there is no rule to strip it. Conforming parsers (sax-js,
libexpat) preserve it.

This was already fixed on `master`/`0.9.x` as a side-effect of the large
DOCTYPE rewrite in PR #498 (22k lines). This PR is the minimal,
non-breaking backport for the maintained `0.8.x` line.

### How

`parseInstruction` builds a substring that already excludes `?>`, so `$`
anchors immediately before it. The `\s*` before `$` was greedily
consuming any trailing whitespace from PI data before passing it to
`domBuilder.processingInstruction`. Removing it — while keeping `*?` on
the data group to minimise diff — is the complete fix.

```js
// before
source.substring(start, end).match(/^<\?(\S*)\s*([\s\S]*?)\s*$/)
// after
source.substring(start, end).match(/^<\?(\S*)\s*([\s\S]*?)$/)
```

Five existing snapshots in
`test/xmltest/__snapshots__/not-wf.test.js.snap` were updated: they
captured the old buggy behaviour (trailing space stripped from
XML-declaration-like PIs in the not-well-formed corpus). The updated
snapshots reflect the now-correct output.

### Scope

Addresses the trailing-whitespace sub-issue from #42, backporting #498
behaviour to 0.8.x.