`substring_from` fails at `rune_start == rune_count(s)`, and mishandles multi-byte runes there

[mororo18]

Context

odin:    dev-2026-08
OS:      Ubuntu 24.10, Linux 6.17.0-19-generic
Backend: LLVM 19.1.1

Expected Behavior

substring_from(s, rune_start) should succeed with an empty result when rune_start == rune_count(s), the same way s[len(s):] is a valid empty slice.

The doc comment says ok reports "whether the rune indexes where in bounds of the original string", which could be read as excluding the one-past-the-end index. But substring already treats that index as in bounds, tests/core/strings/test_core_strings.odin (line) pins substring("Hello", len("Hello"), len("Hello")) == ("", true) , so substring_from should be consistent with it.

Current Behavior

For ASCII strings it returns ok = false instead of succeeding. For strings with multi-byte runes it's worse: it returns a wrong, non-empty, truncated substring instead of an empty one.

Failure Information

Steps to Reproduce

1. Run the snippet below.
2. substring_from("hello", 5) returns ("", false) — expected ("", true).
3. substring_from("héllo", 5) returns ("o", false) — expected ("", true).

Live repro: https://godbolt.org/z/415EzMW1h

Failure Logs

package main

import "core:strings"
import "core:fmt"

main :: proc () {
	fmt.println(strings.substring_from("hello", 5))   // want ("", true), got ("", false)
	fmt.println(strings.substring_from("héllo", 5))    // want ("", true), got ("o", false)
}

Related to #6189, which reported a similar bug in substring, that specific case is already fixed on master, but the same root cause is still present in substring_from and for non ASCII strings.