Fix issues from #9 -- increase select time, replace invalid firmware string

[eode]

The select call has been increased to 0.2 0.5 seconds, which stops the issue originally mentioned in #9 - "Cannot read firmware from device".

A second issue is brought up in #9 - Sometimes the returned firmware string is V3.3 instead of TEMPerX_V3.3. This workaround simply replaces V3.3 with the full TEMPerX_V3.3 string retries when the V3.3 string is encountered.

When the replacement is used, a warning is issued. This goes to stderr, and doesn't interfere with piping the output of the script.

Reproduction

If you haven't applied the fix yet and are attempting to verify the bug's existence:

• use the TEMPer2 device
• 413d:2107 vendor/product ids
• TEMPerX_V3.3
• internal and external temperature sensors
• executing temper.py once a second causes the issue to be much more frequent
• Raspberry pi 3b -- issues may also occur with other hardware, but this is what I'm testing on.

Example

Using xonsh:

import time

while True:
    ./temper.py
    time.sleep(1)

[eode]

Changes

• Firmware read select timeout time was bumped from 0.2 to 0.5 since read failure was still occurring (although rarely).
• Retry by a simple recursion if the V3.3 string is encountered.
  • Retry is probably better than replacing V3.3 with TEMPerXV3.3, since it's possible the bug may be encountered on other TEMPer hardware (that would return something other than TEMPerXV3.3).
  • Since the invalid responses are infrequent, retrying until a valid response occurs is ok, and simple recursion suffices to guarantee a response.
• No longer warns when the issue is encountered.

[nigels-com]

I think that the number of retries ought to be bounded. That's why #16 uses a loop rather than recursion. The other advantage of #16 is that it seems to tolerate a smaller timeout for the select. (Raspberry Pi model B tested).

[eode]

On mine, (raspberry pi 3b) it times out very frequently, using #16.

If having bounded retries (rather than exiting on failure) is preferable, that's fine. But the original issue that brought this up (Issue #9), using #16, is not solved on mine without a select time of at least 0.2 (and even that times out sometimes).

[eode]

#16 now handles the same issues. I'll close this.