Fixed edge case that might lead to Serial beeing stuck in spinlook wait

[MaxMax-embedded]

By creating this pull request you agree to the terms in CONTRIBUTING.md.
https://github.com/Infineon/.github/blob/master/CONTRIBUTING.md
--- DO NOT DELETE ANYTHING ABOVE THIS LINE ---

CONTRIBUTING.md also tells you what to expect in the PR process.

Description

The if condition to store data in the buffer is triggered regardless of the current uart transmit status if tail and head differ:

if(( XMC_USIC_CH_GetTransmitBufferStatus( _XMC_UART_config->channel ) == XMC_USIC_CH_TBUF_STATUS_BUSY )
        || ( _tx_buffer->_iTail != _tx_buffer->_iHead ))

This can lead to an infinte loop in the spinlook if the transmit IRQ is executed in the wrong moment and not new transmission is started. To prevent this a check is added to ensure that a new transmission is started if non is currently in progress by the following code:

  if( XMC_USIC_CH_GetTransmitBufferStatus( _XMC_UART_config->channel ) != XMC_USIC_CH_TBUF_STATUS_BUSY ) 
{
	  XMC_UART_CH_EnableEvent( _XMC_UART_config->channel, XMC_UART_CH_EVENT_TRANSMIT_BUFFER );
	  XMC_UART_CH_Transmit( _XMC_UART_config->channel, _tx_buffer->_aucBuffer[ _tx_buffer->_iTail ] );
	  _tx_buffer->_iTail++;
	  if(_tx_buffer->_iTail >= SERIAL_BUFFER_SIZE)
		  _tx_buffer->_iTail = _tx_buffer->_iTail - SERIAL_BUFFER_SIZE; 
  }

This seems to fix the problem on my setup.

Context
Tested on a XMC4700 Relax Kit using the 3.4.1 Version of the XMC for Arduino Core running 2Mbit Serial1 to a TLE9012 and a 115200 Bit/s Serial to the Host PC

[LinjingZhang]

You are absolutely correct. The old function does not attempt to re-enable the interrupt and send data, which causes the buffer to remain full and the spinlock persists.
If you can simplify the comments, I will merge your PR. thank you very much!

[LinjingZhang]

Could you please simplify the comments, for example by keeping only the “//Reenable IRQ and send Data”