dune-project parser does not work if dune-project has a UTF-8 byte-order mark

Edits

1. Narrowed the issue to just supporting UTF-8 with BOM (ie. supporting the UTF-8 spec). Previously the original problem report was UTF-16 + BOM but as mentioned by @Alizter OCaml does not support UTF-16.

Expected Behavior

In PowerShell on Windows:

echo "(lang dune 3.12)" > dune-project
dune build

I would expect the project to build.

Actual Behavior

File "dune-project", line 1, characters 0-34:
1 | ��(lang dune 3.12)
    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Error: Invalid first line, expected: (lang <lang> <version>)

The reason? Windows conventionally has a byte-order mark in its Unicode files. The built-in PowerShell (or 5.x or less) inserts the BOMs; newer Command Prompts do not.

> with-dkml file dune-project
dune-project: Unicode text, UTF-16, little-endian text, with CRLF line terminators

> with-dkml od -xc dune-project
0000000    feff    0028    006c    0061    006e    0067    0020    0064
        377 376   (  \0   l  \0   a  \0   n  \0   g  \0      \0   d  \0
0000020    0075    006e    0065    0020    0033    002e    0031    0032
          u  \0   n  \0   e  \0      \0   3  \0   .  \0   1  \0   2  \0
0000040    0029    000d    000a
          )  \0  \r  \0  \n  \0
0000046

Specifications

• Version of dune (output of dune --version): 3.12.1
• Operating system (distribution and version): Windows 11 with PowerShell 5.1

Start by reproducing the issue with the PowerShell commands in the report on Windows, using the stated dune 3.12.1 behavior. Trace the dune-project parser from that failure and verify that a UTF-8 BOM is accepted while preserving the existing expected first-line validation.