bad_keywords.txt: Add Nigerian phone number pattern

[tripleee]

As discussed here: http://chat.stackexchange.com/transcript/message/34997711#34997711

234/235 TP (one NAA): https://metasmoke.erwaysoftware.com/search?body=%28%5E%7C%5B%5EA-Za-z0-9_%5D%29%5C%2B234%5B0-9%5D%7B8%2C%7D%28%24%7C%5B%5EA-Za-z0-9_%5D%29&body_is_regex=1&commit=Search&feedback=&feedback_filter=naa&reason=&site=&title=&user_rep_direction=%3E%3D&user_reputation=0&username=&utf8=%E2%9C%93&why=

The proposal is Glorfindel's but we agreed I would make the PR. This is somewhat more far-reaching than your routine blacklisting request so I think we want to have this reviewed by the authors.

[tripleee]

Tangentially notice also the spurious second NAA result, what's up with that?

http://chat.stackexchange.com/transcript/message/34997774#34997774

[tripleee]

This SEDE query finds 21 potential FPs in all of Stack Overflow: https://data.stackexchange.com/stackoverflow/query/619660/nigerian-phones-take-2

[ArtOfCode-]

Does this not get caught by the phone number in {} reason?

[tripleee]

Nope; as noted in the very first comment in the chat discussion I linked, that only gets applied to the title.

[ArtOfCode-]

[0-9] can be \d (digit class). Should probably also put a maximum on the repetition operator - 14 is a common max length for phone numbers, so {8,11} would be okay.

[tripleee]

I am aware of no particular reason to prefer \d over [0-9] but sure. (It complicates matters if you want to use Metasmoke's somewhat less capable MySQL regex engine with these patterns.)

[ArtOfCode-]

@tripleee I suggested \d because semantics. It's always good to be semantically correct where we can.

[tripleee]

This was broken; fix in PR484