Ignore CBOR tags 0-5 and 21-23 (head bytes 0xC0-0xC5, 0xD5-0xD7)

[MsfPablo]

Fixes #5315.

Every byte in 0xC0–0xDB is major type 6 (tag), but the switch in parse_cbor_internal() enumerated only 0xC6–0xD4 and 0xD8–0xDB. Head bytes 0xC0–0xC5 and 0xD5–0xD7 fell through to default: and were reported as parse_error.112 in every handler mode, including ignore and store.

This adds the missing head bytes to the tagged-item case list, so they are handled exactly like tags 6–20 already are:

• cbor_tag_handler_t::error (the default) still throws — unchanged.
• ignore skips the tag head and parses the enclosed data item.
• store parses the enclosed item; as with tags 6–20, no subtype is recorded (only the 0xD8–0xDB forms carry one today).

No tag is interpreted: a text string tagged 0 stays a string, 0xC2 over a byte string stays a byte string.

Changes

• include/nlohmann/detail/input/binary_reader.hpp (and the amalgamated header): add case 0xC0–0xC5 and case 0xD5–0xD7 to the tagged-item block.
• tests/src/unit-cbor.cpp: the "Tagged values" section that swept 0xC6..0xD4 now sweeps 0xC0..0xD7, asserting throw in error mode and the original value in ignore/store mode for each head byte.
• docs/.../cbor.md: the "Incomplete mapping" list no longer claims tags 0–5 and 21–23 are unsupported; the tagged-items note now states the range and that tags are skipped, not interpreted.

The "all CBOR first bytes" test needed no change — it parses in the default error mode, where these bytes still throw parse_error.112.

Verification

test-cbor_cpp11 passes (the only failures in my local run are the test cases requiring the downloaded test-data fixtures, which I did not fetch). The reproduction from the issue now works:

{"t":"2026-07-27T00:00:00Z"}

and a sweep of 0xC0–0xD7 confirms every head byte throws in error mode and yields the enclosed value in ignore and store mode.

[nlohmann]

Apologies for the AI-assisted review here — I don't have time for a deeper pass on this one right now, so this is a quick comparison rather than a full re-review.

This duplicates #5331, which has the identical code fix and was opened first. Your docs update here is actually better — it removes the stale "Incomplete mapping" bullets that #5331 left contradicting the paragraph right below it — but this PR is currently missing a Signed-off-by trailer, so the DCO check is failing and will block merge as-is.

Neither PR yet adds the test coverage requested on #5331 (a store-mode test with a binary payload, not just a string, to confirm 0xC0-0xC5/0xD5-0xD7 take the same "unwrap, don't treat as subtype" path as the existing tag range).

Given #5331 is further along (green CI, already reviewed), we'll likely take that one as the base and fold in the docs fix from here. Thanks for the contribution either way — closing this in favor of #5331 once that's updated, unless you'd rather rebase your docs fix onto it yourself.

— posted by Claude Code on behalf of @nlohmann