GDScript function calls have unexpected overhead

### Tested versions

4.2.2 stable

### System information

Linux Vulkan API 1.3.277 Forward Mobile, Cpu Ryzen 4800

### Issue description

Hello, this is my first issue
the code with functions has low performance,
in other languages (rust, python), the same code gives better performance

When assembled, the performance increases, but not much (0,16s-0.17s and 0.6s-0.8s)

this behavior can be considered a bug because the difference is in other languages, but only 3-8% (for rust), 15-24% (for python)

need to be tested on different platforms

```gdscript

extends Node

var iter : int = 1_000_000
var test_strick : float = 0.0
var time : float = 0.0

func sqrt_v1(num:int) -> float:
return num**0.5

func _ready() -> void:
for d:int in range(5):
time = Time.get_unix_time_from_system()
for i:int in range(iter):
test_strick += sqrt_v1(i)
print(Time.get_unix_time_from_system() - time, " =1= ",test_strick)
time = Time.get_unix_time_from_system()
for i:int in range(iter):
test_strick+= i**0.5
print(Time.get_unix_time_from_system() - time, " =2= ",test_strick)
```
output
0.22625708580017 =1= 666666166.458842
0.07334899902344 =2= 1333332332.91772
0.22873187065125 =1= 1999998499.37655
0.07424306869507 =2= 2666664665.8354
0.22937297821045 =1= 3333330832.29431
0.07326412200928 =2= 3999996998.75321
...

### Steps to reproduce

1) Create empty project with 1 node and 1 script
2) paste code
3) run

### Minimal reproduction project (MRP)

N/A

Start by running the supplied GDScript benchmark from the issue and compare the function-call and inlined-loop timings on multiple platforms. Trace the relevant function-call execution path in the engine, then verify that any confirmed overhead is reduced without changing the benchmark's results.

Written by the indexing model from the issue text.