Prevent Paramiko size mismatch errors in utf-8 strings

[sbj-ss]

Length of unicode strings consisting of multibyte characters differs from their size in bytes. Hence uploads of files and templates using unicode characters will fail with messages like

    Failed to upload file, retrying: size mismatch in put!  208 != 190
    Failed to upload file, retrying: size mismatch in put!  208 != 190
    Failed to upload file, retrying: size mismatch in put!  208 != 190
    [172.33.2.57] Command socket/SSH error: OSError('size mismatch in put!  208 != 190',)
    [172.33.2.57] Error: executed 0/1 commands

Using BytesIO will prevent this.

[Fizzadar]

@sbj-ss thank you for the PR, this definitely seems sensible to me! I’ll fix up those test issues as they’re hidden in some of the legacy test stuff and get this merged!

[Fizzadar]

As part of the v3 work I incorporated this change at a lower level which I think better handles these cases automatically:

• auto convert: 6be2aa4
• tests: d8c1205

Usually I hate implicit behavior (lots of v3 work removes it!) but in this case I think this is best since passing StringIO objects in/out of operations is standard and recommended. This ensures that the output file always matches what is requested in get_file_io. Thank you once again for the PR, inspiration and fix @sbj-ss!