Fix pyserial forcing nonblocking serial device on Linux and spurious null bytes

[jzhan357]

On Linux, pyserial forces vmin to 0 if inter_byte_timeout is None, which is the default behavior.
When vmin is 0, reads to the device can return a minimum of 0 bytes, in other words they return immediately if no data is buffered.
This breaks tools that assume reading from the serial device will block until data is found.
See this thread for more details.

In some cases, spurious null bytes are emitted due to break conditions being interpreted as nulls.
This is fixed by enabling BRKINT and NOFLSH, which will prevent break conditions from being interpreted as null bytes.

[jzhan357]

A bit more poking and I've narrowed down the problem to brkint being turned off in some circumstances.
This leads to break conditions being interpreted as 0x00, which is why we see null bytes inserted between messages.
Either brkint or ignbrk need to be set so that this doesn't happen.

[eggroll-bot]

Adding an update here from what was discussed in Slack:

ignbrk is cleared by pySerial when the serial port is opened, so this cannot be set before the serial port is opened by pySerial. brkint cannot be used because it clears the terminal's input and output buffers upon a break condition resulting in the loss of messages when a break condition occurs.

A minimally reproducible example of the spurious nulls can be found here: https://git.ucsc.edu/ucsc-ectf-2023/serial-null-bytes. I'm seeing null characters from break conditions at the beginning and end of the "abcd" message sent by this example when testing with the bridge.

[eggroll-bot]

Some further testing indicates that ignbrk and echo need to be set after the serial port is opened.

[jzhan357]

Enabling echo is causing failures in firmware loading.