Test suite: add --config-dir isolation, --smoke-test alias, and a cold-start/close gate step

Adopted from Andre's RemSound-smoke-test-agent-brief.md - the gaps our pack didn't
already cover:

- --config-dir <folder>: redirect ALL user state (config, profiles, logs, cue
  sounds) to an explicit folder for this process only, applied at the very start of
  Program.Main before the layout migration runs. Lets a test exercise a real build
  without touching the user's live settings (the brief's safety rule 1). Works with
  every command. AppConfig gains SetUserDataDirectoryOverride / an override on
  UserDataDirectory; CommandLine.TryGetConfigDir parses it early.
- --smoke-test / --smoketest: alias for --selftest, matching the brief's vocabulary.
- run-tests.ps1: a cold-start + clean-close smoke (brief baseline steps 3-4) -
  launches the GUI minimized against an isolated --config-dir, confirms it stays up,
  that it used the isolated folder (real settings untouched), and that --close shuts
  it down with no orphan process. SKIPs cleanly if a RemSound instance is already
  running (machine-wide single-instance lock).

Manual + --help updated for both switches.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

src/RemSound.App/Program.cs

@@ -25,6 +25,16 @@ internal static class Program
             return;
         }
 
+        // --config-dir <folder> (test / portable isolation): redirect ALL user state - config,
+        // profiles, logs, cue sounds - to an explicit folder for THIS process only. Applied first,
+        // before the layout migration and sound consolidation below read or write the default
+        // location, so a smoke test can run a real build without touching the user's settings
+        // (smoke-test brief, safety rule 1).
+        if (CommandLine.TryGetConfigDir(args, out var configDir))
+        {
+            AppConfig.SetUserDataDirectoryOverride(configDir);
+        }
+
         // SustainedLowLatency tells the GC to avoid full (gen 2) collections while audio is streaming.
         // Gen 0/1 collections still happen but are sub-millisecond; the long pauses that were causing
         // the receiver to fall behind in clusters of 4-5 underruns at a time were almost certainly