More queue fetcher fixes/improvements

[philbudne]

Fixes to address issues found in latest A/B comparison runs with scrapy/batch fetcher:

created indexer/requests_arcana.py:

• creates requests Session object for maximal scrapy compatibility
• sets headers directly (avoid creating Connection: header)
• add Accept: header (say we prefer text!)
• nasty hackery to make OpenSSL v3 more tolerant of old versions of TLS

worker.py, storyapp.py: create State Enum for orderly shutdown of Pika thread to avoid story loss!

tqfetcher.py, sched.py:

• raise minimum interval to 5 seconds
• deal with (russian?!) sites that don't return Content-Type header!!
• add ConnStatus.THROTTLE to respond to HTTP 429 status
• add FinishRet for more/better logging of slot state (to debug/test throttling)
• add tunables: --initial-interval-seconds --initial-interval-seconds
• default number of worker threads with goal of keeping 2/3 of all cores busy
• add max_delayed_per_slot parameter (1/4 of prefetch) so that one site can't gum up processing
• requests.Session is a context handler, so use with statement
• improve request average handling to be closer to scrapy
• add comments
• return & log num_delayed from Scoreboard.get_delay

[philbudne]

I've pushed updates to address these