feat: run tests via gotestsum to report the number of tests per testing type

[ergousha]

Closes #301.

What

Runs all test* tasks through gotestsum instead of plain go test, so every run ends with a summary line such as:

DONE 42 tests in 12.345s

Because each testing type (unit, integration, component, e2e) already runs as its own task/matrix job, this yields the per-type test counts asked for in #301.

How

• build/task.yml
  • New vars: GOTESTSUM_BIN, GOTESTSUM_ENABLED (default true), GOTESTSUM_FORMAT (default testname), GOTESTSUM_VERSION.
  • New gotestsum-install task, following the existing gta-install/osv-scanner-install pattern. It is a no-op when GOTESTSUM_ENABLED is false.
  • The test task now picks gotestsum --format <format> -- or go test and passes the existing flags unchanged.
• scripts/package-version-updater.sh - registers GOTESTSUM_VERSION so the weekly updater keeps it current.
• README.md / CLAUDE.md - document the new override variables.

Opting out

includes:
  remote:
    taskfile: ...
    vars:
      GOTESTSUM_ENABLED: "false"

Verified locally

• task remote:test-cicd on a scratch module: DONE 3 tests in 1.374s.
• Failing test still fails the task (exit code propagates through gotestsum).
• task remote:coverage unaffected; -coverprofile is passed through and the threshold check still works.
• GOTESTSUM_ENABLED=false falls back to the previous go test output.
• The taskfile-sorted-units and taskfile-without-empty-lines checks pass.

Out of scope

Aggregating the four counts into a single cross-job summary (artifact upload + $GITHUB_STEP_SUMMARY) is deliberately left out; it can follow up if there is demand.