docs: fix spammer latency CSV example name

[UfukNode]

Noticed the spammer latency docs describe generated files as tx_latency_<yyyymmdd>_<hhmmss>.csv, but the analysis examples used a different tx-latency-... name.

Updated the examples to use the same filename format that the spammer actually writes.

Testing:

• git diff --check
• python3 scripts/test_tx_latency_report.py

[osr21]

Verified against the source and this checks out completely:

• The spammer writes tx_latency_{}.csv with chrono::Utc::now().format("%Y%m%d_%H%M%S") — and it does so at both construction sites in spammer.rs (~line 164 and ~511), so the underscore/timestamp format is the only format the binary can ever produce. The old tx-latency-1234567890.csv examples (hyphens + what looks like a unix epoch) match neither the separator nor the timestamp scheme — likely a leftover from an earlier iteration of the tracker.
• Nice touch reusing tx_latency_20260310_151123.csv verbatim from the README's own "Output" section (line ~438): the doc now shows one consistent example filename from generation through analysis, so a reader can literally trace the same file across sections.
• The remaining data.csv placeholders in the subcommand examples are correctly left alone — those are deliberately generic, and tx_latency_report.py takes whatever --csv path it's given (no filename-pattern assumptions in the script), so nothing else in that section needed touching.

Small observation on why this survived: the stale examples were internally consistent (all three identical), which is exactly the kind of wrong-but-coherent doc that passes casual review. The fix inherits the doc's real defense — the example now derives from the format string the code uses, so a future rename of the CSV pattern will at least have a greppable tx_latency_ prefix linking code and README.

Approving — three-line docs fix, verified against both code paths that write the file.