Drop whitespace-only text nodes between block elements

[trsdn]

Fixes #33

Pretty-printed HTML puts a newline between block elements, and that newline is its own text node. It was collapsed to a single space and appended, so <p>one</p>\n<p>two</p> converted to one\n\n two with a stray leading space on the second block. Anything that pastes formatted HTML hits this, since almost every source indents its markup.

The part that makes this non-trivial

The obvious fix, discarding whitespace-only text nodes, is wrong: between inline elements the very same node is a real word separator, so <em>a</em> <em>b</em> would become *a**b*.

The distinguishing signal turns out to already be in the output buffer. Every block element finishes by appending a newline, so if the buffer ends with a newline the whitespace is layout and can be dropped; otherwise it is inline and is kept as a single space. That also handles leading whitespace at the very start of a document, where the buffer is still empty.

Verification

Full suite: 86 passed, 0 failed, 1 skipped. The remaining skip is #32, which is still open. The inline-spacing cases are covered by the existing tests, which is what pinned down that the naive fix would have regressed them.

Base branch

Stacked on fix/34-code-block-language (PR #51), which is itself stacked on the test project in PR #41. Merge order is #41, then #51, then this.

[trsdn]

Superseded by #53. These commits are already contained in that branch, so this PR would merge nothing on its own. Collapsing the stacked converter PRs into one avoids the base-rewriting problem that made the stack unmergeable.