Avoid too many characters cached in buffer

[huweiATgithub]

Issue Description

Recently, I've encountered an issue that the launch test (added with add_ros_test with a launch.py file) hangs indefinitely.
After investigation, I found that it was stuck at the getvalue call of a StringIO buffer:

launch/launch/launch/actions/execute_local.py

Lines 391 to 396 in f3e9cfe

	def __flush_buffers(self, event, context):
	line = self.__stdout_buffer.getvalue()
	if line != '':
	self.__stdout_logger.info(
	self.__output_format.format(line=line, this=self)
	)

The root cause is that the __on_process_output method stores output in the buffer but does not process it, resulting in a huge buffer to be getvalue on (and thus hangs):

launch/launch/launch/actions/execute_local.py

Lines 378 to 385 in f3e9cfe

	for line in buffer:
	if line.endswith(os.linesep):
	logger.info(
	self.__output_format.format(line=line[:-len(os.linesep)], this=self)
	)
	else:
	last_line = line
	break

This happens because one of my noisy program prints to stdout without a line separator, causing the buffer to accumulate.

Proposed Improvement

This PR addresses the issue by ensuring the buffer does not store too many characters, which I think it is always not wrong.

Other Changes

1. I also change the os.linesep to \n since from the doc of StringIO, the StringIO by default uses \n as newline. Otherwise, writing lines with EOL='\n' in Windows will stop line from being processed.
2. Remove the break in the else clause; this is unnecessary, it will be the last line in the buffer if it does not end with the line separator.

[fujitatomoya]

@mjcarroll can you assign yourself to this? (i do not have power...)