docs(build,test): align README and DEVELOPMENT with the real build (#72)

[luongnv89]

Summary

Closes #72 (F-DOCS-001, F-DOCS-002) — the last issue of the modernization epic #75.

The docs had drifted from the tree across four recently merged PRs (#91, #92, #93, #94). Every factual claim in this PR was verified by running the command or reading the current source; nothing was inferred, and nothing from the issue body (which quotes a MODERNIZATION_PLAN.md that no longer exists) was trusted.

Approach

Three classes of drift were fixed:

1. Broken build commands. The manual compile in docs/DEVELOPMENT.md omitted flows.c and failed at link (undefined reference to 'flows_create', flows_attach, flows_destroy, flows_print_top). The GDB recipe compiled a single translation unit and used pre-subcommand argv. docs/TESTING.md's two unit-test compile lines pointed at the wrong paths (test_config.c instead of tests/test_config.c) and omitted config.c. All four now run verbatim, exit 0, warning-free.

2. Stale prerequisites and build facts. libconfuse-dev was listed as a prerequisite in README.md and docs/DEVELOPMENT.md though config.c implements its own INI parser and the link line is -lmmt_core -ldl -lpcap. The docs now also describe the -Wall -Wextra warning gate (kept in WARNFLAGS so make CFLAGS=... cannot disable it) and the -DMMTREADER_VERSION product-version injection, with file:line citations.

3. Test-count drift — the most-repeated error. Several docs still cited "~194 unit asserts", "13 numbered test groups", and in places "27/27 CLI checks". Totals were re-measured from a real run and standardized everywhere.

Beyond the two files the issue names, the README's sample output was invented (a banner format the binary never printed, tables in a layout it does not use). It is replaced with output captured from ./mmtReader analyze -t smallFlows.pcap -a, including the version object introduced in #93.

Measured totals (standardized across all docs)

Signal	Measured
Numbered test groups	14 (plus sub-groups 2b and 5b)
Unit asserts, groups 5–11 incl. 5b	252 (40 + 9 + 73 + 43 + 38 + 37 + 7 + 5)
CLI integration checks (group 12)	35 / 35
SDK version-gate checks (group 14)	6 / 6
Failures	0
Tolerated skips	1 — live capture on lo (root-gated)

Broken documented commands found and fixed

Command	Where	Failure
manual gcc compile	docs/DEVELOPMENT.md	link error — flows.c missing from the source list
GDB build	docs/DEVELOPMENT.md	compiled only mmtReader.c; run -t ... also skipped the subcommand
gcc ... test_config.c config.c	docs/TESTING.md	test_config.c: No such file — path missing tests/
gcc ... test_parse.c cli/parse.c	docs/TESTING.md	config.h not found; config.c not linked

All four now execute as written.

Changes

File	What changed
docs/DEVELOPMENT.md	Prereqs (libconfuse out, jq in), SDK minimum, real Makefile variables, working manual compile, working GDB recipe, flows.c/cli_mode_t/flows_t added to structure and types, real build/--version output, rewritten Testing section, corrected project layout
README.md	apt line, no-config-library note, make/make test sections with real output, real sample output and JSON, duplicated Usage Examples block removed, -F <seconds> semantics, project structure and docs table
docs/TESTING.md	Full 14-group table with per-suite assert counts, expected totals, tolerated skip, fixed compile commands, test.pcap described accurately
AGENT_ENVIRONMENT.md	13→14 groups, 252 asserts, group 14 row, real compile line, WARNFLAGS/VERSION_DEFS notes, corrected Makefile/test_cli.sh line refs
CLAUDE.md	13→14 numbered test groups
docs/PLAYBOOK.md	Experiment 3 metric 38→37 protocol paths (measured)
docs/CHANGELOG.md	Pointer to the root CHANGELOG.md that holds [Unreleased]
CHANGELOG.md	Documentation entry under [Unreleased]
skills/mmt-reader/references/installation.md	libconfuse removed; SDK ≥ 1.8.0 troubleshooting row added

Test Results

$ make clean && make test
...
Passed: 35 / 35
Failed: 0 / 35
SKIP: live capture on 'lo' unavailable — Couldn't activate device lo: socket: Operation not permitted
All tests passed!
$ echo $?
0

make coverage also verified green (76.53% line / 61.56% branch).

Acceptance Criteria Verification

Criterion	Status	Evidence
Manual build command from docs/DEVELOPMENT.md runs verbatim and links	✅	Extracted the code block programmatically and executed it — exit 0, zero warnings, resulting binary reports mmtReader version: 0.3.0
grep -ri libconfuse README.md docs/ is empty	✅	Exit 1 (no matches)
jq prerequisite from task 0.5 remains	✅	Present in README.md, docs/DEVELOPMENT.md, AGENT_ENVIRONMENT.md, docs/TESTING.md
make clean && make test exits 0, 0 failures, 1 tolerated skip	✅	Output above
README and DEVELOPMENT match actual observed output	✅	Every sample regenerated from live runs; independently re-verified by a review pass that re-ran all commands and opened every cited line range

Notes

• Not fixed, deliberately: .oss-ready/*.md still mentions libconfuse. Those are point-in-time audit artifacts of a past OSS-readiness review, not living documentation; rewriting them would falsify the record. They are outside the acceptance-criteria grep scope.