MarkdownWriter: append unescaped spans in bulk

[tballison]

During profiling after the release of Tika 4.0.0, we found we could improve write speed quite a bit in cases where there weren't a lot of characters that needed to be escaped.

🤖

text() always passes a CharMatcher, so every text literal was written 
to the Appendable one character at a time. Scan for the next character 
that needs escaping and append the clean span in one call instead.

JMH, whole CommonMark spec (SpecBenchmark, JDK 17, -f 3 -wi 8 -i 12)

Benchmark	before	after
renderMarkdownWholeSpec (to StringBuilder)	836.5 ± 34.3 ops/s	956.7 ± 11.3 ops/s	+14%
renderMarkdownWholeSpecToWriter	401.9 ± 2.5 ops/s	837.2 ± 20.1 ops/s	2.08x

Rendering 4.7 MB of plain prose (best-of-22 after warmup, JDK 17)

Target Appendable	before	after
OutputStreamWriter	107 ms	14.5 ms	7.4x
StringWriter	75 ms	12.3 ms	6.1x
StringBuilder	26 ms	12.7 ms	2.0x

Same measurement, 4.0 MB of the spec (escape-dense, closer to a worst case)

Target Appendable	before	after
OutputStreamWriter	54.9 ms	25.8 ms	2.1x
StringWriter	46.1 ms	21.5 ms	2.1x
StringBuilder	24.2 ms	20.8 ms	1.2x

[roxspring]

💭 Worth extracting an append(String, CharMatcher) throws IOException method? Often seems that a comment like this is a signal that a method extraction is available, and the new loop+if looks very self-contained.