supervisor: a worker's ready line is found wherever it sits in its log, not in the first page only (#105)

[linuxnow]

The daemon looked for a worker's ready! only in the first 4 KB of its log. lilv writes a warning line for every entry of LV2_PATH that is no bundle before mod-host says ready, and the whole tests/ directory is 37 such entries, 5334 bytes of them: the marker sat past the page, the daemon waited out ready_timeout_ms and answered resp -503 for every worker given the whole set. An own worker starts because its world is one linked bundle, which writes nothing. That is the second worker of the issue: the pool one, and the one for a URI no bundle names.

log_has_ready now reads the whole log a page at a time, with the tail of each page kept in front of the next so a marker split across two reads is still found.

Tests:

• tests/fake-host takes FAKE_WARN_BEFORE_READY=<n>: n lilv-shaped warning lines on stderr before its ready line.
• a_worker_that_warns_more_than_a_page_before_ready_still_starts (make test-daemon): 64 lines, add answers resp 0 within the ready timeout, and the positive control reads the marker's offset in the worker's log past 4096. Red before the fix with the issue's line: add fake:a 0 -> 'resp -503', wanted 'resp 0'.
• make sabotage gains "the ready marker is looked for in the first page of the worker's log only" (the loop reads one page), and it is caught:
  ok sabotage caught: the ready marker is looked for in the first page of the worker's log only -> a_worker_that_warns_more_than_a_page_before_ready_still_starts red
• make sabotage with the full worker environment on the desk: 187 caught, 0 skipped, one survivor that is the desk's and main's alike (the unapplied nice, sabotage: "a nice the daemon could not apply is not logged" survives where the nice applies #116).
• CI runs the JACK step with LV2_DIR=$PWD/tests, the issue's own reproduce, instead of the one-bundle copy it fell back to. On the desk, make test-jack with that world: 11 ok, among them mod_host_scenarios_pass_unchanged_with_an_lv2_worker and lv2_own_worker_sees_one_bundle_and_replays_after_a_kill, both red before ('add urn:no:such:plugin 4': got 'resp -503', want 'resp -101'; add … 1 pool:l -> 'resp -503'). make test: 171 ok.

Codes: none added, none removed. WORKER_SPAWN (-503) keeps its meaning, the worker is not up; it is no longer answered for a worker that is up behind a noisy world.

Closes #105