Environment files, job variables and steadier runners

Added

• Environment files per stage. Environment files can be stored and managed per stage, and the environments page is split into tabs so each stage's file is reachable on its own.
• Variables in jobs. A job can carry variables, which the runner reads when it executes the job.
• Choose the environment and runner tag for a job. Creating a job now accepts the environment it should run in and the runner tag it should be sent to, and a workflow can record both.

Improved

• A dropped runner connection is handled. Connection handling on the runner was reworked, so a connection lost mid-job is caught and reported instead of going unhandled.
• A flood of log output no longer swamps the database. Log writes are throttled when a job produces output faster than it can be stored, and the writes behind them were reworked to keep the load down.
• Signed-in requests do less work. Sessions of authenticated users are cached, so a request no longer re-hashes the credentials it has already verified.
• A runner's configuration form opens filled in. Opening a runner shows its configuration prefilled with that runner's current details.