[noahkhomer18]

Closes #260.

Both interrupt tests decided when to interrupt with a fixed wall-clock sleep, so a slow or loaded runner could finish the scan before the signal landed. That is the flake class this repo has been clearing out.

Interrupts now fire from an interruptWriter that triggers on the Nth stdout line, so the signal lands while the scan is genuinely running.

• main_e2e_test.go: TestE2EResume fires os.Interrupt after the 3rd result line, replacing the 1s sleep.
• main_signal_unix_test.go: TestE2ESIGTERMExits143 fires SIGTERM after the 1st result line, replacing the 200ms sleep.

Acceptance criterion met: no time.Sleep decides when the interrupt fires in either file.

$ grep -n time.Sleep main_e2e_test.go main_signal_unix_test.go
(no matches)
$ go test -run TestE2EResume -count=20 .
ok  github.com/TMHSDigital/subenum  43.395s
$ go test ./...
19 ok, 0 FAIL

One honest note: main_signal_unix_test.go carries a //go:build unix tag, so on this Windows machine I could not execute that test. It was changed by reading it and mirrors the pattern proven in TestE2EResume; CI on Linux will be the first real run of it.

[noahkhomer18]

Closing this: 536c4b6 on main already fixes #260, and it landed after I branched. Same approach (interrupt on a result line instead of a wall-clock sleep) and the same two files, so this PR is redundant rather than conflicting.