Fix UTF-8 file name encoding for uploads (fixes #3013)

[eliandoran]

Attempting to upload a note containing UTF-8 characters will result in a wrong interpretation of the name. It can easily be tested by creating a file called Helău world.txt and trying to upload it.

Not a client problem, since the payload looks fine:

-----------------------------35722791636796930272418934920
Content-Disposition: form-data; name="upload"; filename="Helău.txt"
Content-Type: text/plain

 Helău world

The parsing gets worse in the server:

{
  fieldname: 'upload',
  originalname: 'HelÄ\x83u.txt',
  encoding: '7bit',
  mimetype: 'text/plain',
  buffer: <Buffer 20 48 65 6c c4 83 75 20 77 6f 72 6c 64 0a>,
  size: 14
}

Seems like there's an entire story regarding why UTF-8 characters are not well parsed:

• Browsers do not always respect a special header to allow UTF-8 file names.
• We are using multer as the Express middleware to process the files for us.
• Someone raised this exact issue in originalfilename error code when originalfilename contain chinese expressjs/multer#1082.
• Seems that multer is using busboy.
• Someone from busboy changed the default from UTF-8 to Latin-7, but allowing the developer to change it.
• The multer library does not change it back to UTF-8 and neither do they allow the developer to change it.
• Someone made a PR in multer to allow fix this (https://github.com/expressjs/multer/pull/1102/files), but it hasn't been merged since Jun 2nd.

The change on the server side is rather trivial: enforce UTF-8 encoding. This fixes the issue and does not seem to introduce any regressions. Tested locally and it now parses the name correctly.

[eliandoran]

@zadam , if you're not comfortable with enforcing the UTF-8 encoding (since it's somewhat of an upstream issue), we could try to upgrade to the non-LTS version of Multer, but from what I understand it breaks compatibility with some older versions.