Default html charset for `stack test --coverage` reports on windows should not be cp437

The charset of of html files produced by `stack test --coverage` depends on the OS/locale. E.g. on Ubuntu 16 all files will be marked with `charset=UTF-8`, while on a Win10 machine they will be `charset=CP437` (which mangles any character outside this charset; editing manually the charset to `utf8` fixes the problem). I did not find any settings that would override the charset/provide custom css/etc.

### Steps to reproduce
1. Take any project with a non-empty test suite
1. Cover in that test suite a file with non-ascii characters inside
1. Use Win10
1. Run `stack test --coverage`
1. Open the produced html report for that file
1. The charset would be set to `CP437`, making non-ascii part of the source code non-readable

### Expected
A setting in a config file/cli option that would put the charset of my choice; or at least `charset=UTF-8` everywhere regardless of the OS.

### Actual
`charset=CP437` where there is no reason to put this charset

### Stack version
1.5.1 (latest after `stack upgrade`); OS=Win10

### Method of installation

* Official binary, downloaded from stackage.org or fpcomplete's package repository