Fix byte offsets of lines returned by LogcatProcessor.tail

[kwy404]

tail() reads the file backwards in 64KB blocks. For every block except the one at the start of the file, the offset of the first complete line did not count the newline after the partial remainder, so all lines in that block got a position one byte too early.

In practice this means that for logcat files over 64KB, a line logged after now() and read back with tail() compares as earlier than the now() marker. The added test covers that case, and offsets from tail() now match those from get_lines().

[xpconanfan]

Nice catch! It seems like can compute current_offset directly inside the if remaining > 0: / else: branches above so we don't need to check if remaining > 0: twice?

          if remaining > 0:
            remainder = split[0]
            lines_chunk = split[1:]
            current_offset = remaining + len(remainder) + 1
          else:
            remainder = b''
            lines_chunk = split
            current_offset = 0

[kwy404]

Thanks, that reads better. Applied in 2d63b39, the offset is now set in each branch and the second check is gone.