tsdb.unescape() is too naive with escaped slashes

[goodmami]

I encountered this bug while doing type checks, not with actual data, but then again it wouldn't trigger an error so it may have just not been noticed.

I recently replaced the regex-based substitution to escape and unescape TSDB strings with simple str.replace() calls, which were about 3-4 time faster. This works well enough for escaping, but for unescaping it causes issues because of the multiple calls. Specifically, an escaped \ followed by a regular s or n would look like an escaped \s or \n:

\\s    replace \\ with \
\s     replace \s with @
@

Switching the order of the replaces wouldn't help, because \\s would become \@ instead of \s. Instead, unescaping needs to return to some sequential low-level parser. Maybe a simple for-loop with if-statements will not be as slow as re.sub().