Restart a crashed language server per --restart never|always|N

The process is supervised: on exit after a successful initialize it is
spawned again with exponential backoff (500ms to 30s), open documents are
re-sent as didOpen before the new session is announced, and requests caught
by the crash wait for it and retry. A crash more than a minute after the
previous one starts a fresh streak; reaching the streak limit (default 3)
ends restarting for good, after which every request fails with an LspError
saying so. Restarts and give-ups are logged and forwarded to the MCP client.

Writes to the child's stdin go through a PassThrough so a broken pipe ends
the session instead of surfacing as an unhandled rejection.

test/fixtures/workspace.ts

@@ -1,11 +1,14 @@
 import { fileURLToPath } from "node:url";
 import { Effect, FileSystem, Layer, Path } from "effect";
-import { LanguageServer } from "../../src/lsp/LanguageServer.ts";
+import { LanguageServer, type Options } from "../../src/lsp/LanguageServer.ts";
 
 const server = fileURLToPath(new URL("./server.ts", import.meta.url));
 
 /** A temporary workspace with the given files, served by the fixture language server. */
-export const workspace = (files: Record<string, string>) =>
+export const workspace = (
+  files: Record<string, string>,
+  options?: Pick<Options, "restart">,
+) =>
   Layer.unwrap(
     Effect.gen(function* () {
       const fs = yield* FileSystem.FileSystem;
@@ -18,6 +21,7 @@ export const workspace = (files: Record<string, string>) =>
         command: process.execPath,
         args: [server],
         root,
+        ...options,
       });
     }),
   );