closes #75

This is a large squash commit. For details, see https://codeberg.org/eltrac/weepinbell/pulls/80

- print clear stack trace using timbre when capturing exceptions
- don't pass nil to url/normalize; if url/normalize gets a nil, instead
of NPE, it now returns an empty string
- capture exceptions in processor pipeline to keep it from dying
- http.downloader now checks response status instead of blindly
downloading anything

- some namespaces tests require did not get renamed
- url/extract might throw AssertionError if input contains non-HTTP(S)
urls. it should not check schemes.
- verifier doesn't need to filter non-HTTP(S) urls since they will
return false anyway, and plus it makes localhost urls in tests fail

- URISyntaxException was not caught in url/normalize, causing
normalization failure which leads to verification failure when there is
one invalid URL, even though valid URL exists
- skip non-HTTP(S) urls when verifying

utils: add url utilities for accurate comparison

- add url/normalize taht reconstructs URL, strips unnecessary
components, reorders query string and ensures determinism
- add url/url= that uses url/normalize

receiver: normalize URLs before checking and putting in queue

- use url= and normalize from utils.url
- rename verify-request to check-request to avoid confusion with
verifier.clj

verifier: check if source includes target URL more accurately

- implement url/extract with robinst/autolink-java, which extracts a
vector a URLs from string
- check if source includes target by
  1. extracting all links from source body
  2. normalizing all links before comparing with target link (with url=
func)
- refactor verify-headers -> match-content-types, with a higher order
function and clearer description

retriever, storage: use url= instead of = when comparing urls