dump: walk multiple armored messages in --list-packets (#2036)

[ronaldtse]

Summary

Fixes #2036. The packet dumper now walks all armored messages in inputs that contain more than one, instead of stopping after the first.

Root cause

DumpContextDst::dump() and DumpContextJson::dump() used init_armored_src() which reads exactly one armor block and reports EOF at the -----END ...----- line. Inputs like data/test_stream_key_merge/key-both.asc (public keys in the first block, secret keys in the second) silently dropped everything after the first block.

The fix is the same pattern already used by process_pgp_signatures() and the key loader: rnp::ArmoredSource constructed with AllowBinary | AllowMultiple, then loop on armor.eof() && armor.multiple() → armor.restart().

Behaviour change

Before:

:armored input
:off 0: packet header 0x99... (tag 6, len 397)    ← public key only
...                                                 ← secret-key block silently dropped

After:

:armored input
:off 0: packet header 0x99... (tag 6, len 397)     ← public key
...
:off 0: packet header 0x95... (tag 5, len 1414)    ← secret key now dumped
:off 1417: packet header 0xb4... (tag 13, len 15)  ← secret-block UID
...

Output now matches gpg --list-packets on the same fixture (modulo intentional formatting differences).

Test plan

• test_ffi_dump_multiple_armored_messages (new) — covers both the text and JSON dumpers; asserts Secret key packet / Secret subkey packet appear in the output and "tag":5 / "tag":7 appear in the JSON
• test_ffi_key_dump, test_ffi_key_dump_edge_cases, test_ffi_signatures_dump — no regression
• Manually compared ./rnp --list-packets src/tests/data/test_stream_key_merge/key-both.asc to the gpg dump attached to issue rnp --list-packets dumps fewer packets than gpg --list-packets from the same source #2036 — same number of packets, same types
• CI green

Out of scope

Issue #2036 also mentions a broader audit (vendor subpackets, partial-length packets, PQC messages). This PR is scoped to the specific bug the maintainer identified in the issue thread (multiple armored messages). The broader audit belongs in follow-up PRs and intersects with #13 (Photo ID) for one of the gap categories.

[codecov]

Codecov Report

❌ Patch coverage is 93.93939% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 85.46%. Comparing base (87657a6) to head (9427b81).

Files with missing lines	Patch %	Lines
src/librepgp/stream-dump.cpp	93.93%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2429      +/-   ##
==========================================
+ Coverage   85.44%   85.46%   +0.01%     
==========================================
  Files         125      125              
  Lines       22946    22962      +16     
==========================================
+ Hits        19607    19625      +18     
+ Misses       3339     3337       -2     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[ronaldtse]

Retargeted onto #2439 (json-c → nlohmann): the multi-armor dump walk is now implemented with nlohmann on top of the migration, instead of json-c APIs that this stack removes. The diff here is now a single commit on top of #2439. Merge order: #2439 → this PR (and #2448, which is also stacked on it).

The nlohmann version is also simpler — merging each armored block's packets into one array no longer needs refcount juggling.

[ni4]

LGTM, thanks!