Record spurious failures

Well, just 'Cannot connect to Docker daemon' failures, for starters.

deploy-guides/0002-spurious-failure-reporting.md

+# Deploy instructions for spurious failure logging
+
+Follow the standard set by [0001-eventlog-importing.md](0001-eventlog-importing.md).
+
+## SQL for creating tables and privileges
+
+```sql
+drop type if exists ci_failure_t cascade;
+create type ci_failure_t as enum ('docker');
+;
+drop table if exists ci_failure;
+create table ci_failure (
+    failure_id bigserial primary key,
+    job_id bigint not null,
+    type ci_failure_t,
+    job_date timestamp with time zone
+);
+;
+grant insert, select on ci_failure to ghc_perf;
+grant usage on ci_failure_failure_id_seq to ghc_perf;
+grant select on ci_failure to ghc_perf_web;
+```