MSBuildExecutor should use explicit Charset to read stdout/stderr

Currently, `MSBuildExecutor` uses the default encoding to read stdout/stderr which might change between builds, is not tracked as an input and thus might introduce flakiness.

Initially discussed in https://github.com/gradle/gradle/pull/6728/files/93911c01ef48ea468113cd00c76fad4a23ddbde9#r218812277

Code reference: https://github.com/gradle/gradle/blob/7a50729b3d3855334ba520f274088e489c019ec3/subprojects/ide-native/src/integTest/groovy/org/gradle/ide/visualstudio/fixtures/MSBuildExecutor.java#L162-L169

Start with the referenced lines in subprojects/ide-native/src/integTest/groovy/org/gradle/ide/visualstudio/fixtures/MSBuildExecutor.java and inspect how stdout and stderr are read. Check the related native integration tests, then verify that both streams use an explicit charset and that the tests pass without relying on the machine's default encoding.

Written by the indexing model from the issue text.