Fix watcher not fully paginating on Init

[clux]

• fixes watcher on 0.92 does not get all objects during Init #1524
• extends the mock test to cover more than one page (test as is here would fail on main)

will cherry-pick to a 0.92.1 branch and release from there
cannot believe i missed this :(

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 75.1%. Comparing base (f553c4e) to head (42178d1).

Additional details and impacted files

@@           Coverage Diff           @@
##            main   #1525     +/-   ##
=======================================
+ Coverage   75.1%   75.1%   +0.1%     
=======================================
  Files         78      78             
  Lines       6980    6991     +11     
=======================================
+ Hits        5238    5249     +11     
  Misses      1742    1742             

Files	Coverage Δ
kube-runtime/src/watcher.rs	42.3% <100.0%> (+0.3%)	⬆️
kube/src/mock_tests.rs	98.3% <100.0%> (+0.4%)	⬆️

[clux]

Annoyingly this also "appears to undo" parts of the memory benefits that i benchmarked for 0.92.. will have to update the post later, but so far, even the starting memory (5m in after init) is doing better than 0.91, but it is more situational

controller	0.91	0.92 with page_size 50	this pr page_size 50	this pr page_size 20
2k pod labeller	45M	9M	23M	20M
ks controller	64M	51M	64M	60M

given that these controllers have also had about a week to run 0.92 it's also clear that these leak much more memory than before, and in context of the bug, this also kind of makes sense; the reflectors were not full at the beginning - and the gains i saw by measuring at the start would not last anyway.

maybe there are some more optimizations to do, but it's more important to not have a broken release out here.

[Danil-Grigorev]

/lgtm

:)

[clux]

released in https://github.com/kube-rs/kube/releases/tag/0.92.1 from 0.92.x branch (cherry-picked in this commit).