Fix IfElseRuntime for Frida: ensure both runtime branches initialize to restore ASAN

[Rohankaf]

Description

This PR fixes issue #3460, where IfElseRuntime caused Frida ASAN instrumentation to stop working.

Root Cause

IfElseRuntime only executed init() and deinit() for the selected branch.
However, Frida-based runtimes — especially AsanRuntime — rely on init_all() to register:

• ASAN shadow memory mappings
• Frida instrumentation hooks
• coverage mapping callbacks
• comparison logging hooks

When asan was wrapped inside IfElseRuntime, these hooks were never registered, causing ASAN to silently fail.

Fix

This PR updates the implementation so that:

• init() calls init_all() for both if_runtimes and else_runtimes
• deinit() calls deinit_all() for both runtime branches

pre_exec() and post_exec() remain conditional (intended behavior).

This ensures that all runtimes correctly register their instrumentation during initialization, even when wrapped in IfElseRuntime.

Result

This now works exactly as expected:

IfElseRuntime::new(move || Ok(true), tuple_list!(asan), tuple_list!());

And behaves identically to using asan directly, restoring ASAN crash detection and shadow checks.

This fully resolves #3460.

---

Checklist

• I have run ./scripts/precommit.sh and addressed all comments
• I verified the modified code compiles
• The fix is isolated to the IfElseRuntime implementation

[s1341]

Thansk for the PR, but please do NOT use AI to write your PR messages.

[Rohankaf]

hey @s1341 Note: CI failures appear unrelated to this PR.
All errors are coming from libafl_libfuzzer_runtime and tinyinst_simple, neither of which are touched by this change. The modified code affects only crates/libafl_frida/src/helper.rs. These failures are already present on main , what should i do ?

[domenukk]

This fix makes no sense, if the runtime is not enabled it doesn't need to be initialized, right?