feat(#67)!: infallible Matter::to_qb64 in core (text encoding without stream)

[joeldsouzax]

Summary

Resolves the DevX / layering wart in #67: encoding a Matter-family primitive (Verfer, Diger, Saider, Prefixer, Signer, Cigar) to qb64 text no longer requires the stream feature.

• core: Matter<C> gains infallible to_qb64() -> String and to_qb64b() -> Vec<u8>, co-located with the existing decode path (MatterBuilder::from_qualified_base64). Mirrors the sibling Indexer::to_qb64/Siger::to_qb64 contract — the only failure modes are internal-invariant breaks (corrupt sizage table / mis-sized buffer), unreachable for a validated Matter, handled via unreachable!/assert_eq! (programmer-bug carve-out, not a data path).
• stream: stream::encode::matter_to_qb64 removed; all consumers (3 examples, 1 bench, 1 fuzz test) migrated to the new API. examples/encode_primitive now needs only core, not stream — a direct demonstration of the issue's goal.
• serder: to_qb64_string / identifier_to_qb64_string (and the internal seal_to_json / matters_to_json_array) are now infallible (-> String / -> Value); the misplaced SerderError::Qb64Encoding(ParseError) and its dead companion SerderError::Encoding(FromUtf8Error) are removed. Fixes the error-domain smell where a read-path ParseError surfaced on a write path.

Breaking changes (MINOR under 0.x — see CHANGELOG)

• stream::encode::matter_to_qb64 removed → use Matter::to_qb64b.
• serder::primitives::{to_qb64_string, identifier_to_qb64_string} return String instead of Result<String, SerderError>.
• SerderError loses the Qb64Encoding and Encoding variants (breaking for downstream exhaustive match).

Test Plan

• nix flake check green — all 11 checks (clippy god-level, rustfmt, taplo, audit, deny, nextest across feature combinations, doctest, wasm, no_std). 1777 tests pass.
• Round-trip tests relocated into core and run without the stream feature: from_qualified_base64(x) ↔ to_qb64b()/to_qb64() across fixed, variable-soft, typed/untyped codes.
• Added ls>0 (lead-byte) and xs>0 (xtra underscore-fill) coverage — mutation-proven: each new test fails under the corresponding one-line mutation.
• Encode/decode verified as exact inverses (matching framing arithmetic, inverted).

Closes #67.

🤖 Generated with Claude Code

[codspeed-hq]

Merging this PR will not alter performance

✅ 21 untouched benchmarks

---

_{Comparing feat/67-matter-to-qb64-in-core (0fe2246) with main (1c62ce8)}