internal/router/scheduler: add serial scheduler, default on this fork

Add a strict one-model-at-a-time scheduler. Requests run in exact
arrival order; at most one runs at a time; switching to a different
model evicts every other running model first so a single model occupies
memory at a time. Unlike fifo it never reorders or batches same-model
requests, and it ignores group/matrix co-residency entirely, making the
single-model guarantee a property of the scheduler rather than the config.

- new Serial scheduler implementing the Scheduler interface
- register "serial" in scheduler.New; default routing.scheduler.use to
  "serial" at config load (fifo still selectable for upstream behavior)
- update config schema, example config, and config defaults tests

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

internal/router/scheduler/scheduler.go

@@ -92,9 +92,14 @@ type Effects interface {
 	StopProcesses(timeout time.Duration, ids []string)
 }
 
-// New returns a Scheduler selected by conf.Routing.Scheduler.Use, configured
-// from conf and bound to the given planner and effects. Currently only "fifo"
-// (the default) is supported.
+// New returns a Scheduler selected by conf.Routing.Scheduler.Use, configured from
+// conf and bound to the given planner and effects. Supported values are "fifo"
+// (throughput-oriented, batches same-model requests) and "serial" (strict
+// one-model-at-a-time, exact arrival order).
+//
+// The deployment default is applied by config loading (LoadConfig sets Use to
+// "serial" when unset). The "" fallback here is the library default and remains
+// "fifo" so callers that build a Config directly keep the original behavior.
 func New(conf config.Config, name string, logger *logmon.Monitor, planner Swapper, eff Effects) (Scheduler, error) {
 	use := conf.Routing.Scheduler.Use
 	if use == "" {
@@ -103,6 +108,9 @@ func New(conf config.Config, name string, logger *logmon.Monitor, planner Swappe
 	switch use {
 	case "fifo":
 		return NewFIFO(name, logger, planner, conf.Routing.Scheduler.Settings.Fifo, conf.Models, eff), nil
+	case "serial":
+		// Serial ignores the group planner: it always evicts every other model.
+		return NewSerial(name, logger, eff), nil
 	default:
 		return nil, fmt.Errorf("unsupported scheduler type: %q", use)
 	}