gh-146073: Add fitness/exit quality mechanism for JIT trace frontend

[cocolato]

Reopen: #147966

• Issue: Improving trace quality by tracking "fitness" and "exit quality" #146073

[cocolato]

Build configure: ./configure --enable-experimental-jit --enable-optimizations

The new bench result with higher initial value and lower frame penalty:

+---------------------+----------+-----------------------+
| Benchmark           | baseline | fitness               |
+=====================+==========+=======================+
| raytrace            | 236 ms   | 209 ms: 1.13x faster  |
+---------------------+----------+-----------------------+
| pyflate             | 262 ms   | 250 ms: 1.05x faster  |
+---------------------+----------+-----------------------+
| xml_etree_iterparse | 60.3 ms  | 58.2 ms: 1.04x faster |
+---------------------+----------+-----------------------+
| json_dumps          | 6.29 ms  | 6.08 ms: 1.03x faster |
+---------------------+----------+-----------------------+
| float               | 44.2 ms  | 42.8 ms: 1.03x faster |
+---------------------+----------+-----------------------+
| xml_etree_parse     | 94.2 ms  | 91.4 ms: 1.03x faster |
+---------------------+----------+-----------------------+
| richards            | 15.4 ms  | 15.0 ms: 1.02x faster |
+---------------------+----------+-----------------------+
| chaos               | 39.9 ms  | 39.1 ms: 1.02x faster |
+---------------------+----------+-----------------------+
| nbody               | 48.1 ms  | 47.1 ms: 1.02x faster |
+---------------------+----------+-----------------------+
| spectral_norm       | 51.8 ms  | 50.8 ms: 1.02x faster |
+---------------------+----------+-----------------------+
| deltablue           | 2.11 ms  | 2.07 ms: 1.02x faster |
+---------------------+----------+-----------------------+
| go                  | 80.6 ms  | 79.3 ms: 1.02x faster |
+---------------------+----------+-----------------------+
| xml_etree_process   | 44.0 ms  | 43.3 ms: 1.01x faster |
+---------------------+----------+-----------------------+
| xml_etree_generate  | 65.6 ms  | 64.9 ms: 1.01x faster |
+---------------------+----------+-----------------------+
| regex_compile       | 92.4 ms  | 91.5 ms: 1.01x faster |
+---------------------+----------+-----------------------+
| json_loads          | 17.0 us  | 17.1 us: 1.01x slower |
+---------------------+----------+-----------------------+
| telco               | 4.96 ms  | 5.02 ms: 1.01x slower |
+---------------------+----------+-----------------------+
| fannkuch            | 215 ms   | 220 ms: 1.02x slower  |
+---------------------+----------+-----------------------+
| Geometric mean      | (ref)    | 1.02x faster          |
+---------------------+----------+-----------------------+

[cocolato]

It appears that the current parameters do not yet guarantee runtime safety; I will continue to work on fixes and optimizations.