pydfu: fix windows stdout.flush issue on python 3.6

[andrewleech]

There appears to be an issue with cpython >= 3.6 sys.stdout.flush() on windows as used in pydfu.

File: firmware.dfu
    b'DfuSe' v1, image size: 1751517, targets: 1
    b'Target' 0, alt setting: 0, name: "ST...", size: 1751232, elements: 2
      0, address: 0x08000000, size: 15736
      1, address: 0x08020000, size: 1735480
    usb: 0483:df11, device: 0x0000, dfu: 0x011a, b'UFD', 16, 0x61194f1d
Writing memory...
0x08000000   15736 [                         ]   0% Traceback (most recent call last):
  File "src\micropython\tools\pydfu.py", line 548, in <module>
    main()
  File "src\micropython\tools\pydfu.py", line 539, in main
    write_elements(elements, args.mass_erase, progress=cli_progress)
  File "src\micropython\tools\pydfu.py", line 454, in write_elements
    progress(elem_addr, 0, elem_size)
  File "src\micropython\tools\pydfu.py", line 485, in cli_progress
    sys.stdout.flush()
OSError: [WinError 87] The parameter is incorrect

I think it's the same as described in: pytest-dev/py#103

Their workaround appears to be fairly long winded and I'm not sure it's really necessary here (https://github.com/pytest-dev/pytest/pull/2462/files)

I've found it works just fine to catch and ignore the error on the flush line.

Tested:
Windows 10 x64 1803 (Build 17134.228)
Python 3.6.4 amd64

[dpgeorge]

It looks ok to me.

But did you try sys.stdout.buffer.flush() or sys.stdout.buffer.raw.flush() instead? It's a long shot but might work.

[andrewleech]

No, never thought of that. I'd be more wary of either of them not working on other platforms.

I don't think it's particularly buffered on windows anyway, the progress bars seem to work fine anyway.

[dpgeorge]

Ok, merged in a270364