[trace] Moving quantization before truncation of spans

[ryplo]

This is to ensure that strings (ex. query strings) stored in certain fields, such as resources are still parseable and are quantized properly.
Tested in staging and prod

[palazzem]

@ufoot I think it's safe to ship that huge improvement together with the distributed sampling. If you don't have strong opinion, I'm going to add that one to our 5.18.2 milestone.

[ufoot]

Left a few remarks about testing.

[ufoot]

I'm generally speaking in for this, the only thing I'm worried about is Quantize eating lots of CPU when the resource is very long. We have a protection against this, the watchdog is going to drop traces at pre-sampling stage if this happens. I think it's fine to ship this as is, but let's keep an eye on performance, not saying we should not do it, just, pay attention to side effects.

[ryplo]

Good point! Will keep that in mind

[ufoot]

I think we lack a test here, something that would test that say UPDATE blogpost SET comment='<...something possibly huge, at least longer than 5000 char...>' WHERE blogpost_id=42; and see how that ends. We should end up with the right quantization, which shows the WHERE clause. My understanding is that the purpose of this PR is to treat that case, because before, truncation would have removed the WHERE clause before anything else. I might have missed that test, or an equivalent, if that's the case, just point me to it.

[ryplo]

Discussed with Christian - decided to add a test mimicking agent behaviour to test joined effort of quantizer and truncator.