feat(sched): cancel in-flight actions when a build fails

Only new spawns were held back after a failure, so every action already running
was allowed to finish even though its output would be discarded. The scheduler
now sends SIGTERM to the remaining children on the first failure, and reports
them separately rather than counting a process it killed as a failure of its
own. Partial outputs are removed and their cache entries dropped, so the next
run starts clean.

Under -k the existing behaviour is kept, since keeping going is the point.

On a graph with three six-second generators alongside one translation unit that
fails immediately, -j4 now returns in 0.12 s rather than 6.03 s.

bin/meowc.ml

@@ -205,7 +205,7 @@ let execute p ~names =
         end)
       order;
     Printf.printf "\n  %s\n" (Style.dim (Printf.sprintf "%d actions, nothing run" !n));
-    (b, { Sched.built = 0; cached = 0; failed = 0 })
+    (b, { Sched.built = 0; cached = 0; failed = 0; aborted = 0 })
   end
   else begin
     let cache = Cache.load (Filename.concat p.tc.builddir ".meowc-cache") in
@@ -223,12 +223,14 @@ let execute p ~names =
 let summarise (r : Sched.result) elapsed =
   if not fl.quiet && not fl.dry then begin
     let text =
-      if r.failed > 0 then Style.red (Style.plural r.failed "action" ^ " failed")
+      if r.failed > 0 then
+        Style.red (Style.plural r.failed "action" ^ " failed")
+        ^ (if r.aborted > 0 then Style.dim (Printf.sprintf ", %d cancelled" r.aborted) else "")
       else if r.built = 0 then Style.dim "nothing to do"
       else Printf.sprintf "%d built%s" r.built (Style.dim (Printf.sprintf ", %d cached" r.cached))
     in
     Printf.printf "%s  %s%s\n"
-      (if r.built + r.failed = 0 then "" else "\n")
+      (if r.built + r.failed + r.aborted = 0 then "" else "\n")
       (Style.pad 48 text) (Style.dim (duration elapsed))
   end;
   if r.failed > 0 then exit 1