Use T.decodeUtf8 + BS.readFile instead of T.readFile

[Bodigrim]

As documentation (https://hackage.haskell.org/package/text-2.0/docs/Data-Text-IO.html#v:readFile) explains, T.readFile is almost always a mistake. It's highly unlikely that files we try to read with T.readFile are in a locale encoding and not in UTF-8. Moreover, with text-2.0 fmap T.decodeUtf8 . BS.readFile is up to 10x faster than T.readFile.

[pepeiborra]

Sounds like we need to deprecate T.readFile and introduce T.readFileUtf8 ?

[jneira]

It's highly unlikely that files we try to read with T.readFile are in a locale encoding and not in UTF-8

But it is more than possible that some users, specially in windows are using a local encoding, even without being aware of it.
I guess this pr would break hls for them
so imo at least a warning should be added in the changelog (will do myself)
Ideally a LSP alert warning should be added when loading a file if we detect its encoding is not utf8, will open an issue about

[jneira]

Opened #2641

[jneira]

lgtm, thanks!

[Bodigrim]

But it is more than possible that some users, specially in windows are using a local encoding, even without being aware of it.
I guess this pr would break hls for them

No, it's actually vice versa. This PR fixes HLS for people, whose locale is not UTF-8. GHC mandates that source files are in UTF-8 only, but before this PR HLS would try to use a locale encoding instead.