Uniform `--format json` on every read command

[KKranthi6881]

Problem

Some read-side commands (datalex validate, datalex info, datalex diff) support --format json, but the flag isn't uniform — some commands only emit human output, and the JSON shapes vary across commands that do. This makes it harder to script DataLex into CI and other tools.

Goal

Every read-side datalex subcommand should accept --format {human,json} (default human). When --format json is passed:

• Output is a single top-level JSON document on stdout.
• No log/warning/banner text on stdout; route those to stderr.
• The schema is stable and documented — document each command's JSON shape in docs/cli.md.

Suggested approach

1. Inventory the read commands (search packages/cli/src/datalex_cli/ for add_parser / subparsers).
2. Factor a tiny helper — emit_result(result: Any, *, format: str) — that pretty-prints or JSON-dumps.
3. Flip each command over to the helper.
4. Add one test per command that asserts --format json output parses and includes expected keys.

Acceptance criteria

• datalex validate --format json → diagnostics array.
• datalex info --format json → project summary.
• datalex diff --format json → structured diff (reuse diff.py's internal dataclasses).
• datalex packages list --format json → resolved-packages list.
• No stdout contamination when --format json is active.

Why this is a good first issue

Small surface per command, easy to verify, no core engine changes. Gives you a tour of the CLI layer and forces clean separation of presentation from logic.

[snakefood3232]

The issue is that read commands like validate and info have inconsistent JSON output, making scripting difficult. I'd handle this by first auditing the CLI parsers, then implementing a shared emit_result helper to standardize the JSON format and suppress logs to stderr. I can have a PR up within a day—I've built uniform CLI output for several Python tools.