refactor: load settings once per run and reuse shared helpers

The application caches the settings snapshot and reloads it only after
an update through the store, so output and prompt helpers no longer
read and parse config.yml and hosts.yml several times per command.
canPrompt and chooser replace the inline prompt-policy checks.

Checks are tallied by one services.Tally, terminal.Plural replaces the
two app copies, refKind.article/noun replace the comment group's own
words, list commands use servicePage, and web URLs are built by
repositoryURL and webURL in browser.go.

Signed-off-by: bircni <bircni@icloud.com>

internal/services/checks.go

@@ -32,6 +32,21 @@ func (c Check) Pending() bool { return c.State == "pending" }
 // Passed reports whether the check finished without blocking a merge; Gitea counts skipped as success.
 func (c Check) Passed() bool { return c.State == "success" || c.State == "skipped" }
 
+// Tally counts checks by outcome.
+func Tally(items []Check) (failing, pending, passed int) {
+	for _, item := range items {
+		switch {
+		case item.Pending():
+			pending++
+		case item.Passed():
+			passed++
+		default:
+			failing++
+		}
+	}
+	return failing, pending, passed
+}
+
 // Checks returns latest contexts for the PR head in its base repository, as Gitea does.
 func (s Pulls) Checks(ctx context.Context, n int64) ([]Check, error) {
 	sha, err := s.checkHead(ctx, n)
@@ -126,16 +141,8 @@ func (s Pulls) WatchChecks(ctx context.Context, n int64, interval time.Duration,
 			}
 			previous = items
 		}
-		pending := len(items) == 0
-		failed := false
-		for _, item := range items {
-			if item.Pending() {
-				pending = true
-			} else if !item.Passed() {
-				failed = true
-			}
-		}
-		if failed || !pending {
+		failing, pending, _ := Tally(items)
+		if failing > 0 || (pending == 0 && len(items) > 0) {
 			current, err := s.checkHead(ctx, n)
 			if err != nil {
 				return nil, err