Remove infinity support for tcp listener backlogs

[CoolCat467]

This PR makes due on my comment in 6f9ab95, removing support for math.inf as an argument for the tcp listener backlog, leaving only int | None as valid options.

[codecov]

Codecov Report

Merging #2842 (39380e8) into master (d7c5c64) will increase coverage by 0.00%.
The diff coverage is 100.00%.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #2842   +/-   ##
=======================================
  Coverage   99.18%   99.18%           
=======================================
  Files         115      115           
  Lines       17585    17594    +9     
  Branches     3140     3142    +2     
=======================================
+ Hits        17442    17451    +9     
  Misses         99       99           
  Partials       44       44           

Files	Coverage Δ
trio/_highlevel_open_tcp_listeners.py	100.00% <100.00%> (ø)
trio/_highlevel_ssl_helpers.py	100.00% <ø> (ø)
trio/_tests/test_highlevel_open_tcp_listeners.py	98.51% <100.00%> (+0.05%)	⬆️

[A5rocks]

While I get using int | None to work around flaws in the type system, I think there should at least be a deprecation period.

OH I should read the docs first. Oops. Deprecation not needed. Ignore me.

[njsmith]

Let's just keep accepting inf even if it's not documented or listed as a valid type, as a little backcompat thing? no reason to break existing code and it's not really any extra maintenance burden.

[CoolCat467]

Let's just keep accepting inf even if it's not documented or listed as a valid type, as a little backcompat thing? no reason to break existing code and it's not really any extra maintenance burden.

Added warning for inf but still accepts it as of daada74

[jakkdl]

I'd kinda like making this consistent across the whole codebase with corresponding DeprecationWarnings*, but this seems like a good thing on it's own.

* probably making it a very long deprecation period, since this mostly just is about the typing limitation (though I also do think it's much easier to specify None than having to import math.inf). Or not even raising a warning and just continue to accept math.inf without issue indefinitely, but add support for None (where that's not already the case) and specify the signature as only int|None.

[A5rocks]

nit: feel free to merge regardless.