Fix bug in SingleValueReplacer

[adewaleo]

SingleValueReplacer.process_request properly decodes request.body if its type is byte.

Current implementation of process_request plays poorly with vcr.request.body setter as it encodes strings back to bytes.

The following assumes python 3.

\>>> bytes("foo", 'utf-8')
b'foo'
\>>> str(bytes("foo", 'utf-8'))
"b'foo'"

`simply casting the bytes with string returns the string repr of the byte.`
`This is not a byte type.`

\>>> str(bytes("foo", 'utf-8'), 'utf-8')
'foo'

`decodes byte into string`

\>>> str(bytes("foo", 'utf-8')).encode('utf-8')
b"b'foo'"
`byte contents includes byte literal repr,`

`this is what happens in request.body's setter because of incorrect value`

\>>> str(bytes("foo", 'utf-8'), 'utf-8').encode('utf-8')
b'foo'

`proper byte contents`

cc: @lmazuel

See this line in Azure/azure-cli#9431

[adewaleo]

I don't think this approach is cross platform though, as str(encoding='utf-8') was added in python 3

[adewaleo]

bytes is an alias for str from python 2.6 onwards in python 2, so this won't NameError