refactor(sema): seal sema.zig as editor indexing only (A8.1)

Remove the last compiler dependency on sema as semantic truth and stop
publishing as-you-type sema diagnostics from the LSP.

- core.zig: drop dead `Compilation.analyze()`, the `sema_result` field,
  and the sema->diagnostics merge; drop the now-orphaned sema import.
  The CLI pipeline (parse -> resolveImports -> generateCode) never called
  analyze(), so this removes only dead code.
- lsp/server.zig: rename `analyzeAndPublish` -> `refreshEditorIndex` and
  delete its sema-diagnostic publish (and the now-unused `semaToLspDiags`).
  The editor index (doc.sema) is still refreshed for nav/refs/completion/
  tokens. On-save/on-open diagnostics still come solely from the canonical
  compiler pipeline in `runProjectCheck` (unchanged).
- Document sema as an editor-indexing API (doc.sema field comment).

Intended behavior change: as-you-type sema diagnostics no longer publish;
on-save canonical diagnostics are the sole source. CLI compile output and
the 361-example suite are unchanged (361/0, zero snapshot churn).

src/lsp/document.zig

@@ -23,7 +23,9 @@ pub const Document = struct {
     version: i64,
     /// AST root for this file only (not merged).
     root: ?*sx.ast.Node,
-    /// Sema results for this file (references are relative to this source).
+    /// Editor index for this file — symbols/references/types for navigation,
+    /// completion, and hover (references are relative to this source). Not a
+    /// diagnostic source; see `sema.zig` module doc.
     sema: ?sx.sema.SemaResult,
     /// Last successful sema (preserved across parse failures for completions).
     last_good_sema: ?sx.sema.SemaResult = null,