build: reduce overhead of fuzz testing

[andygrove]

Which issue does this PR close?

N/A

Rationale for this change

CI runs the PR builds with the default native scan implementation specified in the workflow, for example:

- name: "Spark 3.5.6, JDK 17, Scala 2.13"
  java_version: "17"
  maven_opts: "-Pspark-3.5 -Dspark.version=3.5.6 -Pscala-2.13"
  scan_impl: "native_comet"

In CometFuzzTestBase, we were running each test once for each scan implementation. This was redundant.

What changes are included in this PR?

• CometFuzzTestBase uses the default scan impl.
• Update the Linux CI workflow to have one run in auto mode. This was missing.

How are these changes tested?

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 60.06%. Comparing base (f09f8af) to head (fbc8516).
⚠️ Report is 876 commits behind head on main.

Additional details and impacted files

@@             Coverage Diff              @@
##               main    #3257      +/-   ##
============================================
+ Coverage     56.12%   60.06%   +3.93%     
- Complexity      976     1427     +451     
============================================
  Files           119      172      +53     
  Lines         11743    15926    +4183     
  Branches       2251     2631     +380     
============================================
+ Hits           6591     9566    +2975     
- Misses         4012     5031    +1019     
- Partials       1140     1329     +189     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.