fix: restore missing zprivate/replace function files

[IronCrest-sudo]

Bug

stringlib:util/replace (and datalib:core/lib/string/replace, which wraps it) always returned the original string unchanged and a return value of 0, regardless of input.

Root cause

zprivate/replace/ (8 files) was deleted in 43b8bc5 ("refactor: replace old datapack with the new modular structure") along with the rest of data/stringlib/. When fd4d15c ("Organize code") later restored data/stringlib/ — util/, zprivate/concat/, zprivate/find/, zprivate/split/, etc. — zprivate/replace/ was not included.

util/replace.mcfunction still calls stringlib:zprivate/replace/main, check_word_start_loop, and the reversed/ equivalents. Minecraft's function command no-ops silently on an unknown function reference rather than erroring, so the datapack loaded without any "failed to load function" errors — the bug was silent.

Fix

Restored the 8 files from 43b8bc5^ (last known-good state, one commit before the deletion):

zprivate/replace/check_word_rest.mcfunction
zprivate/replace/check_word_start_loop.mcfunction
zprivate/replace/main.mcfunction
zprivate/replace/replace.mcfunction
zprivate/replace/reversed/check_word_rest.mcfunction
zprivate/replace/reversed/check_word_start_loop.mcfunction
zprivate/replace/reversed/main.mcfunction
zprivate/replace/reversed/replace.mcfunction

Verification

Tested against a live 26.2 server via RCON, not just static analysis:

• datalib:core/lib/string/replace with string="Hello World", find="World", replace="Everyone" → output "Hello Everyone" (previously "Hello World", unchanged)
• Regression check: removed zprivate/replace/ again on the live server → bug reproduced ("Hello World"). Restored → fixed again ("Hello Everyone"). Confirms these files are the actual root cause, not a coincidental fix.
• stringlib:util/replace called directly with n=1 on "aXbXcX" → "a-bXcX" (only first match replaced, n param works)
• stringlib:util/replace called directly with n=unset on "aXbXcX" → "a-b-c-", return value 3 (correct match count, was always 0 before)
• stringlib:util/find unaffected (regression check on separate function)

Not in scope

datalib:core/lib/string/replace.mcfunction itself does not propagate a return value from stringlib:util/replace (its last line is a tellraw, no return). stringlib:util/replace's own return value is correct after this fix. Leaving the wrapper as-is since it's unrelated to the reported bug — happy to open a separate PR if wanted.