Raise the CI database statement timeout to 10 minutes

[roed-math]

Summary

Raise the default PostgreSQL statement timeout used by the test tox environment
from 120 seconds to 10 minutes.

Motivation

The two-minute timeout introduced in #7181 successfully prevents unbounded
queries, but it also acts as a sharp performance threshold against the shared
devmirror database. The completeness test has now crossed that threshold in
multiple runs even though the underlying logical result is correct (e.g.
https://github.com/LMFDB/lmfdb/actions/runs/31268853079/job/93204410081, which
canceled the null-data probe for an intrinsically impossible nf_fields
query while it was waiting on the timeout).

Ten minutes still catches a pathological statement far earlier than the
60-minute GitHub Actions job timeout while leaving substantially more headroom
for transient load.

Scope

This PR intentionally changes only tox.ini. The expensive completeness query
and CI matrix concurrency are addressed separately.

Verification

• tox config -e tests resolves the default to statement_timeout=10min.
• An exported PGOPTIONS value still overrides the default (checked with PGOPTIONS='-c statement_timeout=0' tox config -e tests).
• git diff --check passes.

🤖 Generated with Claude Code

[roed-math]

Note on the close/reopen in the timeline: the first CI run hit a transient psycopg.errors.ConnectionTimeout while collecting tests on the devmirror modular_curves shard (two full 12-job matrices were hitting devmirror simultaneously — the load situation #7185 caps). Since this branch has no admin rights to rerun jobs, closing and reopening retriggered CI on the same commit; the retriggered run is fully green (13/13 checks).