Speed up parse() ~7x with literal pre-filters (output unchanged)

[Malkiz223]

Fixes #32.

parse() spent almost all of its time running heavy regexes over the whole log - even where there is nothing to find. The idea is simple: before each regex, do a quick plain-substring check (in / str.find) and only run the regex where a match is actually possible. The regex still decides everything - the check just skips the parts that can't match, so the output stays the same.

log	before	after	speedup
142 KB	13.7 ms	1.8 ms	7.6×
1.1 MB	108 ms	15 ms	7.3×
10 MB	896 ms	126 ms	7.1×
1.1 MB (many error/retry/redirect lines)	98 ms	14 ms	7.2×

What changed (common.py gets a few dicts of marker substrings, the rest is in scrapylogparser.py):

• re_search_final_match - can take a substring now. Lines that don't contain it are skipped before the regex runs.
• extract_latest_matches - same for the 9 patterns, and it takes the last Scraped from straight from the end (rfind) instead of walking every item in the log.
• extract_log_categories - if a category (ERROR, WARNING, etc.) isn't in the log, skip it. If it is, find its lines by substring and cut them out, instead of scanning the whole log with a regex per category.
• extract_datas - the page/item-count regex now runs only on lines that contain " pages (at".
• extract_shutdown_reason - the SIGTERM regex runs only if the line contains "Received SIG".
• extract_stats_dumped - the stats block (always at the end) is found with a plain search, and the regex starts from there instead of byte 0.

Correctness: the existing tests pass unchanged, and everything parse() returns matches the original byte-for-byte (checked on real logs and on a synthetic \n/\r\n log that has every category, SIGTERM, offsite/duplicate, and "gave up retrying"). Passes the project's linter (flake8). These same changes are already running in our production, and everything works correctly.