docs: publish RAM + CPU footprint comparison (Rust vs Python)

Whole-process real-names benchmark: peak RAM ~0.43GB vs ~1.27GB (~3x less),
CPU time 3.0s vs 45.5s (~15x less), measured with /usr/bin/time -v.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Author: jqueguiner

README.md

@@ -73,6 +73,21 @@ indicative only.
 
 \* Not identical logic — Rust port vs upstream faker `first_name`.
 
+### Resource usage
+
+Whole-process footprint of the real-names benchmark (load the 1.43M-name bank,
+then 3M ops), measured with `/usr/bin/time -v`:
+
+| Metric | Python | Rust (release) | Ratio |
+|---|---:|---:|---:|
+| Peak RAM (RSS) | ~1.27 GB | ~0.43 GB | ~3.0× less |
+| CPU time (user+sys) | 45.5 s | 3.0 s | ~15× less |
+| Wall clock | 38.3 s | 3.0 s | ~13× less |
+
+RAM is dominated by the in-memory name index (per-country + global pools).
+Both hold the full 1.43M-row dataset resident; Rust's compact `HashMap` +
+`String` storage uses ~3× less than the Python `dict` structures + interpreter.
+
 Reproduce:
 
 ```bash

rust/README.md

@@ -97,7 +97,9 @@ f.first_name_real(Some("JP"), Gender::Female);        // weighted female JP name
 | `first_name_like_real` | 3.79 M ops/s | 0.84 M ops/s |
 | bank load (1.43M rows) | 2.3 s | 8.9 s |
 
-~4–5× faster on identical logic. See top-level `README.md` for the full table.
+~4–5× faster on identical logic. Whole-process footprint (load bank + 3M ops):
+peak RAM ~0.43 GB vs Python ~1.27 GB (~3× less), CPU time 3.0s vs 45.5s
+(~15× less). See top-level `README.md` for the full tables.
 
 ## Grammatical number agreement