schedule,shared: move concurrency 429 limits into scheduler code (#849)

- make concurrency limiting the scheduler.Scheduler's responsibility
- eliminate the separate concurrency limit middleware 
- move concurrencyLimit logic into scheduler.FIFO to maintain backwards compatibility
- add HTTPError from #834 

Updates #834

internal/router/matrix.go

@@ -6,7 +6,6 @@ import (
 	"github.com/mostlygeek/llama-swap/internal/config"
 	"github.com/mostlygeek/llama-swap/internal/logmon"
 	"github.com/mostlygeek/llama-swap/internal/process"
-	"github.com/mostlygeek/llama-swap/internal/router/scheduler"
 )
 
 type Matrix struct {
@@ -27,10 +26,10 @@ func NewMatrix(conf config.Config, proxylog, upstreamlog *logmon.Monitor) (*Matr
 	// Build a process for every model in the config. Any model can run alone
 	// even if it is not part of a set; this mirrors proxy.NewMatrix.
 	processes := make(map[string]process.Process, len(conf.Models))
-	base := newBaseRouter("matrix", conf, processes, proxylog,
-		func(name string, logger *logmon.Monitor, eff scheduler.Effects) scheduler.Scheduler {
-			return scheduler.NewFIFO(name, logger, swapper, conf.Routing.Scheduler.Settings.Fifo, eff)
-		})
+	base, err := newBaseRouter("matrix", conf, processes, proxylog, swapper)
+	if err != nil {
+		return nil, fmt.Errorf("creating base router: %w", err)
+	}
 
 	for mid, modelCfg := range conf.Models {
 		procLog := logmon.NewWriter(upstreamlog)