fix(#4053): quote decimal-shaped frontmatter scalars for spec YAML readers (#4165)

* fix(frontmatter): quote decimal-shaped scalars so a spec YAML reader preserves them

A decimal phase identifier written to STATE.md frontmatter (e.g.
`current_phase: 22.10`) was emitted BARE, because `scalarNeedsDoubleQuoting`
only asks whether a value can OPEN a plain scalar — which `22.10` can. A
YAML-spec reader (js-yaml, the statusline, any external tool) then reloads bare
`22.10` as the float 22.1, colliding with `22.1` and dropping the trailing zero.
gsd's own tolerant line-scanner (`extractFrontmatter`) round-trips the raw text
and so hid the defect; a spec reader does not.

Fix: `reconstructFrontmatter`'s general scalar path now also quotes numeric-
looking strings that are not plain all-digit integers (decimals, exponents,
sexagesimal, hex/oct/bin) via `generalScalarNeedsNumericQuoting`, reusing the
existing `YAML_NUMERIC_RE`. Every all-digit string — integer counts, phase
numbers, and leading-zero fixtures like `02` — stays bare, so the state-rebuild
idempotency baseline and the rest of the state corpus are unchanged. This also
quotes `gsd_state_version: 1.0` on write, which matches the authoritative
STATE.md template (`src/state.cts` already emits it quoted).

Regression test drives the real write path and asserts, via js-yaml, that
`22.1` and `22.10` no longer collide and read back string-typed; guards that
integers and free-text stay unquoted.

Fixes #4053

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YBicDMJyh3AH56ZFbUsyxC

* chore(changeset): add Fixed fragment for #4053

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YBicDMJyh3AH56ZFbUsyxC

* docs(frontmatter): trim the generalScalarNeedsNumericQuoting comment

Cut the over-long doc block down to the essential why and drop the inline
comment that repeated it.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011ZKeSj55VakqQajtoBgCTC

* docs(test): drop the #4053 explanatory comments from the touched tests

The assertions speak for themselves; remove the added narrative comments.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011ZKeSj55VakqQajtoBgCTC

* fix(#4053): correct the trade-off comment, changeset PR number, and cover every claimed numeric form

Review follow-ups (trek-e):
- The doc comment claimed a plain integer round-trips harmlessly. That is
  false for leading-zero values (`02` -> 2, `017` -> 17 under js-yaml). Rewrite
  it to state the real, deliberate trade-off: all-digit strings stay bare
  because zero-padded ids (`plan: 01`, `phase: 02`) are the pervasive GSD
  convention and quoting them all is the blanket quoting #4053 asked to avoid;
  the loss is padding not identity (`02` and `2` normalize to the same phase,
  `22.1` and `22.10` do not).
- Changeset carried the auto-closed draft's number (4151); correct to 4165.
- Test exponent, hex, octal, binary and sexagesimal forms through js-yaml, and
  pin the leading-zero trade-off so the documented behaviour is asserted.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qc7VN4zTpTSDTS9JXM2cFB

---------

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: Tom Boucher <trekkie@nomorestars.com>

src/frontmatter.cts

@@ -913,6 +913,23 @@ function agentScalarNeedsDoubleQuoting(s: string): boolean {
   return false;
 }
 
+/**
+ * #4053 — Quote a numeric-looking scalar that is not all-digit (`22.10`,
+ * `1.0`, `1e3`, `0x1F`) so a spec YAML reader keeps it a string: bare `22.10`
+ * reloads as the float 22.1 and collides with `22.1`, a different phase.
+ *
+ * All-digit strings stay bare as a deliberate, scoped trade-off — not a
+ * safety guarantee. A leading-zero value (`02`, `017`) also mis-parses under
+ * a spec reader (to 2, 17). It is left unquoted because zero-padded ids
+ * (`plan: 01`, `phase: 02`) are the pervasive GSD convention, quoting them
+ * all is the blanket quoting #4053 asked to avoid, and the loss is padding
+ * rather than identity: `02` and `2` normalize to the same phase; `22.1` and
+ * `22.10` do not.
+ */
+function generalScalarNeedsNumericQuoting(s: string): boolean {
+  return YAML_NUMERIC_RE.test(s) && !/^\d+$/.test(s);
+}
+
 function reconstructFrontmatter(obj: Frontmatter): string {
   const lines: string[] = [];
   // #3257: read the full-line-comment channel (set by parseGuardedYamlRegion when comments
@@ -978,12 +995,12 @@ function reconstructFrontmatter(obj: Frontmatter): string {
         } else {
           // eslint-disable-next-line @typescript-eslint/no-base-to-string
           const sv = String(subval);
-          lines.push(`  ${subkey}: ${sv.includes(':') || sv.includes('#') || scalarNeedsDoubleQuoting(sv) ? `"${escapeDoubleQuotedScalar(sv)}"` : sv}`);
+          lines.push(`  ${subkey}: ${sv.includes(':') || sv.includes('#') || scalarNeedsDoubleQuoting(sv) || generalScalarNeedsNumericQuoting(sv) ? `"${escapeDoubleQuotedScalar(sv)}"` : sv}`);
         }
       }
     } else {
       const sv = String(value);
-      if (sv.includes(':') || sv.includes('#') || sv.startsWith('[') || sv.startsWith('{') || scalarNeedsDoubleQuoting(sv)) {
+      if (sv.includes(':') || sv.includes('#') || sv.startsWith('[') || sv.startsWith('{') || scalarNeedsDoubleQuoting(sv) || generalScalarNeedsNumericQuoting(sv)) {
         lines.push(`${key}: "${escapeDoubleQuotedScalar(sv)}"`);
       } else {
         lines.push(`${key}: ${sv}`);