Lock the DASH7 leakage SNR-curve experiment before it runs

[emrefbulut]

What changed

Adds docs/leakage-real-snr.md: the locked command, seeds, wait conditions, artifact names, and hour estimate for repeating the synthetic SNR curve on DASH7. The curve has not been measured.

Why

ROADMAP Now item 1 is still the n=15 SNR sweep on a real capture (artifacts/leakage_real_runs.json does not exist). The stride half is already published. Starting that grid without locking n, without waiting for IQFORGE_DASH7 and an idle parity gate, or by using the script default SNR list, would either contend with the gate or reproduce the ceiling table the pilot already ruled out.

How it was verified

# artifacts absent
Test-Path artifacts/leakage_real_runs.json artifacts/leakage_real_table.md
# False False

# recipe has no machine paths; IQFORGE_DASH7 / --source only
# SDR CLIs and matching PnP devices: none
# grid not started

No training run. No download. Turkish docs left untracked.

Conventions it touches

• 3 Do not claim what you did not measure — the recipe states the table does not exist
• 5 Do not enlarge a sample after seeing the result — n=15 locked before any run
• 8 Nothing session- or machine-specific is committed — env var / --source only