feat(ucase): add default case mapping package

ucase/README.md

@@ -0,0 +1,121 @@
+# ucase: Unicode Default Case Mapping & Folding
+
+Implementation of the Unicode **default (language-insensitive) case algorithms** — case folding, lowercasing, uppercasing, and titlecasing — plus the `Cased`/`Case_Ignorable` properties and default caseless matching.
+
+## Why `ucase` and not `uaxNN`/`utsNN`?
+
+The rest of this repository follows the `uaxNN`/`utsNN` naming because those packages implement *numbered* Unicode annexes. **Case mapping is not a numbered annex.** It is specified in the core Unicode Standard, **Section 3.13 "Default Case Algorithms"** (with the caseless-matching definitions D144–D147), backed by the UCD data files. There is no "UAX #21." The package is therefore named for its function: `ucase` ("Unicode case").
+
+## Overview
+
+Section 3.13 defines four operations and two derived properties:
+
+| Operation | Function | Notes |
+|-----------|----------|-------|
+| `toCasefold` | `ToFold` | Full folding (`CaseFolding.txt` status C + F) |
+| `toLowercase` | `ToLower` | Full (1:many); applies the `Final_Sigma` context |
+| `toUppercase` | `ToUpper` | Full (1:many); e.g. `ß` → `SS` |
+| `toTitlecase` | `ToTitle` | Titlecase the first cased char of each word, lowercase the rest |
+
+Properties: `IsCased` (D135) and `IsCaseIgnorable` (D136). Default caseless match (D144): `CaselessMatch`.
+
+## Installation
+
+```bash
+go get github.com/SCKelemen/unicode/v6/ucase
+```
+
+## Quick Start
+
+```go
+package main
+
+import (
+	"fmt"
+
+	"github.com/SCKelemen/unicode/v6/ucase"
+)
+
+func main() {
+	fmt.Println(ucase.ToLower("ΟΔΟΣ"))                 // οδος   (final sigma)
+	fmt.Println(ucase.ToUpper("straße"))                // STRASSE
+	fmt.Println(ucase.ToTitle("hello, world"))          // Hello, World
+	fmt.Println(ucase.ToFold("effort"))                  // effort
+	fmt.Println(ucase.CaselessMatch("Straße", "STRASSE")) // true
+}
+```
+
+## API Reference
+
+#### String operations (full, 1:many)
+- `ToFold(s string) string` — default full case folding (toCasefold).
+- `ToLower(s string) string` — default full lowercasing, including `Final_Sigma`.
+- `ToUpper(s string) string` — default full uppercasing.
+- `ToTitle(s string) string` — default titlecasing (see word boundaries below).
+
+#### Rune operations (simple, 1:1)
+- `SimpleFold(r rune) rune` — simple case fold (status C + S). *Distinct from `unicode.SimpleFold`.*
+- `SimpleLower(r rune) rune`, `SimpleUpper(r rune) rune`, `SimpleTitle(r rune) rune`.
+
+#### Properties & matching
+- `IsCased(r rune) bool` — Cased property (D135).
+- `IsCaseIgnorable(r rune) bool` — Case_Ignorable property (D136).
+- `CaselessMatch(a, b string) bool` — default caseless match (D144): `ToFold(a) == ToFold(b)`.
+
+#### Metadata
+- `UnicodeVersion` — the Unicode version of the generated tables (`"17.0.0"`).
+
+## Semantics & Scope
+
+### Titlecasing word boundaries
+`ToTitle` titlecases the first cased character of each word and lowercases all
+other cased characters. A word begins at the first cased character following the
+start of the string or a character that is neither cased nor case-ignorable
+(whitespace, punctuation, symbols). Case-ignorable characters (apostrophes,
+combining marks) do not break words, so `"don't stop"` → `"Don't Stop"`. This is
+a documented, self-contained boundary definition; it is not the full
+[UAX #29](https://www.unicode.org/reports/tr29/) word segmentation.
+
+### Locale tailoring — out of scope (future work)
+Only the **default, language-insensitive** algorithm is implemented. The
+Turkish/Azeri (`tr`/`az`) and Lithuanian (`lt`) tailorings in
+`SpecialCasing.txt` are **not** applied. Consequently the only casing context
+condition the default algorithm exercises is `Final_Sigma`; the others
+(`After_Soft_Dotted`, `More_Above`, `Before_Dot`, `After_I`) appear exclusively
+on language-tagged lines and will be added with locale tailoring.
+
+### Caseless matching
+`CaselessMatch` is the **default** caseless match (D144). It does **not** apply
+canonical normalization, so it is not the **canonical** caseless match (D145),
+which requires NFD. To compare strings that may differ in canonical composition,
+normalize both (e.g. with
+[`uts15`](https://github.com/SCKelemen/unicode/tree/main/uts15)) before calling.
+
+## Conformance
+
+Built from Unicode 17.0.0 UCD data:
+
+- [`UnicodeData.txt`](https://www.unicode.org/Public/17.0.0/ucd/UnicodeData.txt) — simple mappings (fields 12/13/14)
+- [`SpecialCasing.txt`](https://www.unicode.org/Public/17.0.0/ucd/SpecialCasing.txt) — full + conditional mappings
+- [`CaseFolding.txt`](https://www.unicode.org/Public/17.0.0/ucd/CaseFolding.txt) — case folding (C/F/S/T)
+- [`DerivedCoreProperties.txt`](https://www.unicode.org/Public/17.0.0/ucd/DerivedCoreProperties.txt) — `Cased`, `Case_Ignorable`
+
+- `TestDataFileConsistency` checks the generated tables against the committed UCD files (every default mapping).
+- `TestOfficialConformance` downloads the official files and re-verifies (skipped when offline or in `-short`).
+
+## Code Generation
+
+```bash
+cd ucase
+go run generate_ucase.go
+```
+
+The generator is version-pinned to Unicode 17.0.0, produces deterministic
+(code-point-sorted) output, formats with `go/format`, and skips
+language-sensitive `SpecialCasing.txt` lines. Zero external dependencies.
+
+## References
+
+- The Unicode Standard, **Section 3.13 "Default Case Algorithms"** — <https://www.unicode.org/versions/latest/>
+- [UAX #44: Unicode Character Database](https://www.unicode.org/reports/tr44/)
+- [UCD 17.0.0](https://www.unicode.org/Public/17.0.0/ucd/)