Support non-Markdown text files in `zest-dev dump`

Problem

zest-dev dump archives Markdown files from a spec directory, but non-Markdown text artifacts such as SQL scripts are not preserved. A downstream archive workflow may then see the archive issue as complete and delete the source directory, silently losing repository-only text files.

Desired behavior

• Recursively archive UTF-8 text files in addition to Markdown, including at least .sql, .json, .yaml, .yml, .txt, and shell scripts.
• Preserve each file's relative path and complete contents.
• Use a representation that safely handles content containing Markdown fences.
• Treat unsupported or binary files as an observable failure; do not report a successful dump while silently omitting them.
• Do not signal dump completion until every file in the spec directory has either been preserved or caused an explicit failure.
• Keep existing Markdown dump behavior compatible.

Acceptance criteria

1. A fixture containing spec.md and a nested validation.sql is dumped with both relative paths and byte-equivalent text contents recoverable from the archive issue.
2. A spec containing an unsupported binary file exits non-zero and identifies the unsupported path.
3. A partial upload or failed comment creation exits non-zero rather than reporting success.
4. Tests cover mixed Markdown/text directories and text containing triple-backtick sequences.
5. The supported text-file and unsupported-file behavior is documented.

Start at the zest-dev dump command and trace the existing Markdown archive flow, then inspect how its tests create fixtures and upload or create comments. Extend the coverage for mixed nested text files, fence-containing content, unsupported binaries, and failed uploads. Done means all supported files are recoverable with relative paths and exact contents, while unsupported or partial operations fail visibly and the supported-file behavior is documented.