fix: Spray-and-pray hotfixes to avoid exploding event sizes in store

[untitaker]

ref getsentry/sentry#12375

We are running into the problem that running parts of the store normalizer on the same payload will bloat events to ridiculous sizes. Make sure that:

• We don't store the same error twice
• We don't store too large original_value

[untitaker]

For the curious third-party spectator, here are some fun metrics we recorded in datadog:

When the blue line goes up, we turned on rust renormalization.

The purple line is the largest event size recorded at that time interval. Subject to datadog's sampling/magic/bugginess. We have larger events for sure. Anyway, there's a spike that shouldn't be there.