Meta: Upgrade ESMeta to v0.8.1

[arcde40]

This PR updates the version of ESMeta to v0.8.0 with the following key updates:

• Update to ES2026 es-meta/esmeta#345
• Support JSON productions es-meta/esmeta#353
• Add support for terms "finite Number" and "non-zero finite Number" es-meta/esmeta#347
• fix math/number inconsistency es-meta/esmeta#348
• Support precedence-aware compound conditions es-meta/esmeta#358

This version removes SnapToInteger from esmeta-ignore.json, but newly adds five algorithms:

• BigIntBitwiseOp
• DecimalEscape[0,1].CapturingGroupNumber
• INTRINSICS.Function.prototype.bind
• TimeString
• TimeZoneString

All of them are false alarms in newly supported features, caused by a limitation in the current type inference: integer and range constraints are over-approximated to Math through arithmetic operations. For example, in BigIntBitwiseOp, the arguments to BinaryAnd/BinaryOr/BinaryXor are always 0 or 1 by modulo 2, but inferred as Math rather than Int[0, 1].

[ljharb]

@arcde40 would you please register as a contributor?

[github-actions]

The rendered spec preview for this PR is available as a single page at https://tc39.es/ecma262/pr/3962 and as multiple pages at https://tc39.es/ecma262/pr/3962/multipage .

[jhnaldo]

The esmeta typecheck failure here comes from a type inference bug in ESMeta v0.8.0: integer types were over-approximated to Math, so ten spurious alarms cascaded out of ToAbsoluteIndex, ToClampedIndex, and ArraySpeciesCreate. es-meta/esmeta#366 fixes it, released as v0.8.1, and all ten alarms are gone, so none of them need an esmeta-ignore.json entry.

@arcde40 could you update this PR to v0.8.1? It is three lines:

--- a/.github/workflows/esmeta-installer/action.yml
+++ b/.github/workflows/esmeta-installer/action.yml
@@ -13,8 +13,8 @@ runs:
     - name: set esmeta version and path
       shell: bash
       run: |
-        # v0.8.0
-        echo "ESMETA_VERSION=12e35a2478d9553ebc64d4d79d4b8dd57df973c6" >> $GITHUB_ENV
+        # v0.8.1
+        echo "ESMETA_VERSION=1db8ac10c6fb57d81dbaafbb3648440fa8f8be01" >> $GITHUB_ENV
         echo "ESMETA_HOME=vendor/esmeta" >> $GITHUB_ENV

--- a/esmeta-ignore.json
+++ b/esmeta-ignore.json
@@ -12,7 +12,7 @@
   "GetViewByteLength",
-  "INTRINSICS.Function.prototype.bind",
+  "IncrementModuleAsyncEvaluationCount",
   "Record[SourceTextModuleRecord].ExecuteModule",

INTRINSICS.Function.prototype.bind type-checks on the current base, so its entry is dropped (the failing job asks for that too). IncrementModuleAsyncEvaluationCount replaces it: another false alarm of the same class, where a non-negative count is inferred as Int instead of Int[0+].

Please also retitle the PR and the commit message to v0.8.1.

[arcde40]

Thanks for the review! I've updated ESMeta to v0.8.1 and submitted the contributor registration.

Sorry for the noise earlier. I accidentally left some merge conflict markers in and force-pushed a fix, so it should be clean now.