Immediately commit the last data fragments

Sometimes, the last fragments are written after user call `do_io_close`. This means the writing of the last fragments might be pending when user delays `do_io_close`. Then the `read_while_write` keep a loop to wait for this fragments. This dose great harm for performance .

But if ats writes the last fragment immediately, the RWW reader can found the last fragments and terminate itself entirely without any pending .

https://github.com/apache/trafficserver/blob/6c51b258f90163abc667c9ec2036af76940fbdac/iocore/cache/CacheWrite.cc#L1305-L1312