[ty] Add some missing calls to normalized_impl

[AlexWaygood]

Summary

I hoped this might fix the latest stack overflows on #18659... it doesn't look like it does, but these changes seem like they're probably correct anyway...?

Test Plan

[github-actions]

mypy_primer results

Changes were detected when running on open source projects

jinja (https://github.com/pallets/jinja)
- TOTAL MEMORY USAGE: ~106MB
+ TOTAL MEMORY USAGE: ~97MB

black (https://github.com/psf/black)
- TOTAL MEMORY USAGE: ~129MB
+ TOTAL MEMORY USAGE: ~142MB

alerta (https://github.com/alerta/alerta)
- TOTAL MEMORY USAGE: ~106MB
+ TOTAL MEMORY USAGE: ~117MB

nox (https://github.com/wntrblm/nox)
-     memo fields = ~72MB
+     memo fields = ~66MB

mkosi (https://github.com/systemd/mkosi)
- TOTAL MEMORY USAGE: ~117MB
+ TOTAL MEMORY USAGE: ~129MB
-     memo fields = ~97MB
+     memo fields = ~106MB

tornado (https://github.com/tornadoweb/tornado)
-     memo fields = ~142MB
+     memo fields = ~129MB

paasta (https://github.com/yelp/paasta)
-     memo fields = ~171MB
+     memo fields = ~156MB

altair (https://github.com/vega/altair)
-     memo fields = ~251MB
+     memo fields = ~228MB

rotki (https://github.com/rotki/rotki)
-     memo fields = ~445MB
+     memo fields = ~490MB

scikit-learn (https://github.com/scikit-learn/scikit-learn)
- TOTAL MEMORY USAGE: ~717MB
+ TOTAL MEMORY USAGE: ~652MB

[carljm]

Thank you!

[carljm]

This is the thing I'm least happy with in the current approach, is that it's so easy to just call .normalized(db) on a nested type when it should be .normalized_impl(db, visitor). But I haven't come up with any good ideas for preventing that mistake, because Type::normalized does need to exist, for "external" use -- it just shouldn't be used inside another normalized_impl method.