fix: use f-string in _normalize_token error messages

[moltate]

Fixes #579

The _normalize_token function raised a ValueError with a broken
error message — the format string used {token!r} as a regular string
instead of an f-string, so users saw the literal placeholder instead of
the actual invalid token value.

Changes:

• gemma/gm/text/_sampler.py: convert to f-string
• gemma/research/t5gemma/sampling.py: same fix
• gemma/gm/text/_sampler_test.py: add regression tests

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.