split call_function into 14 inline-never family dispatchers

[danieljohnmorris]

Summary

• call_function was a single ~4,500-line function with ~134 if builtin == ... arms. In debug builds, rustc reserves stack slots for every local in a frame simultaneously even across mutually exclusive branches, so each recursive user-fn call consumed several hundred KiB of stack.
• PR feature: crypto primitives - sha256, hmac-sha256, base64, hex, ct-eq #494 (sha2/hmac crate) and PR feature: --allow-net/read/write/run CLI capability flags #500 (capability flags) both needed band-aids: a per-test 8 MiB thread wrapper and RUST_MIN_STACK=16777216 in CI respectively.
• This PR fixes the root cause by splitting the arms into 14 #[inline(never)] family dispatchers. The new call_function is a thin match router (<100 lines). Per-frame stack pressure is now O(one family's locals) regardless of how many builtins exist.

Repro before / after

Before (debug build, default 8 MiB stack):

ilo "fib n:n>n;<=n 1 n;a=fib -n 1;b=fib -n 2;+a b" fib 30
# => SIGSEGV / thread stack overflow at around depth 22-25

After:

ilo "fib n:n>n;<=n 1 n;a=fib -n 1;b=fib -n 2;+a b" fib 30
# => 832040

What's in the diff

Commit 1 - split call_function into 14 inline-never family dispatchers

• dispatch_map_builtins, dispatch_linalg_builtins, dispatch_math_builtins, dispatch_datetime_builtins, dispatch_basic_builtins, dispatch_list_builtins, dispatch_io_builtins, dispatch_text_builtins, dispatch_fs_builtins, dispatch_json_builtins, dispatch_hof_builtins, dispatch_stat_builtins, dispatch_regex_builtins, dispatch_fft_builtins
• resolve_fn_ref and closure_captures promoted from nested functions inside call_function to module-level items, shared by list and HOF dispatchers
• HOF dispatchers (list, hof) take env: &mut Env because srt/rsrt key-fn and map/flt/fld call call_function recursively
• New call_function is a thin match builtin { ... } router with Len inlined (high-frequency, trivial)
• No functional change to any builtin's behaviour

Commit 2 - regression tests (tests/regression_call_function_stack.rs)

• fib30_no_stack_overflow_vm / fib30_no_stack_overflow_jit
• fold_range_no_stack_overflow_vm / fold_range_no_stack_overflow_jit
• fib30_cross_engine_parity

Commit 3 - examples/deep-recursion.ilo

• fib(10/20/30) and sumto(100/1000) with -- run: / -- out: assertions; exercises all example engines

Test plan

• cargo build clean (no warnings)
• cargo clippy -- -D warnings clean
• cargo test --lib - 3274 passed / 0 failed (AOT tests need libilo.a via symlink)
• cargo test - all integration tests pass including new regression_call_function_stack
• cargo test --test examples_engines - deep-recursion.ilo assertions pass
• fib(30) = 832040 on VM and JIT in debug build, no overflow

Follow-ups

• PR feature: crypto primitives - sha256, hmac-sha256, base64, hex, ct-eq #494 (crypto-primitives) and PR feature: --allow-net/read/write/run CLI capability flags #500 (capability-flags) can remove their band-aids after rebasing onto this
• The RUST_MIN_STACK=16777216 in rust.yml on PR feature: --allow-net/read/write/run CLI capability flags #500 should be dropped in its rebase

[codecov]

Codecov Report

❌ Patch coverage is 55.64516% with 110 lines in your changes missing coverage. Please review.
✅ All tests successful. No failed tests found.

Files with missing lines	Patch %	Lines
src/interpreter/mod.rs	55.64%	110 Missing ⚠️

📢 Thoughts on this report? Let us know!

[danieljohnmorris]

Closing to unstick the merge queue. The keep-both rebase strategy can't handle this PR cleanly (it touches the same dispatch table every other PR appends to, producing broken-brace artifacts on rebase). Will reimplement against current main after the rest of the queue drains.

[danieljohnmorris]

Superseded by fresh-impl PR (in flight on feature/-v2 branch). The original PR's keep-both rebase produced broken Rust that can't be unstuck without manual brace surgery. The v2 PR is a clean reimpl against current main.