ruby `fuzz_prism` never calls `ruby_init()`, crashes on every non-empty input

`projects/ruby/fuzz_prism.cpp` never calls `ruby_init()`, so
`pm_arena_new()` dies in `ruby_xcalloc` on a NULL VM before any input is parsed.

```cpp
if (size == 0)
return 0;
pm_arena_t *arena = pm_arena_new(); // line 37
```

In a CRuby build, Prism’s `xcalloc` is `ruby_xcalloc`, which reads
`GET_VM()->ractor.main_ractor`. The VM pointer is NULL, so the fault address is
`0x558`. The input is unused: every file of one byte or more crashes the same way.
The other Ruby harnesses in this project all call `ruby_init()`; this one does not.

```
==1991560==ERROR: AddressSanitizer: SEGV on unknown address 0x000000000558
==1991560==The signal is caused by a READ memory access.
SCARINESS: 10 (null-deref)
#0 ruby_xcalloc_body /src/ruby/gc.c:5282
#1 ruby_xcalloc /src/ruby/gc.c:5276:34
#2 LLVMFuzzerTestOneInput /src/fuzz_prism.cpp:37:25
SUMMARY: AddressSanitizer: SEGV /src/ruby/gc.c:5282 in ruby_xcalloc_body
```

```sh
python3 infra/helper.py reproduce ruby fuzz_prism crash-00000289082f96f2c1a33d4217daceb834efb03c
```

Any non-empty file behaves the same.

The issue was already reported to ruby, and redirected here.

Found by the CISPA Fandango Team.

Start with projects/ruby/fuzz_prism.cpp, especially LLVMFuzzerTestOneInput near line 37, and compare initialization with the other Ruby harnesses. Reproduce with python3 infra/helper.py reproduce ruby fuzz_prism crash-00000289082f96f2c1a33d4217daceb834efb03c. Done means non-empty inputs no longer crash before parsing.

Written by the indexing model from the issue text.