Improve implementation of MappedWrite

[edmorley]

MappedWrite currently manually appends a single element at a time to its buffer in its io::Write implementation (rather than a split and push several at once approach, or similar):

libcnb.rs/libherokubuildpack/src/write.rs

Lines 107 to 114 in fbf3919

	fn write(&mut self, buf: &[u8]) -> io::Result<usize> {
	for byte in buf {
	self.buffer.push(*byte);
	if *byte == self.marker_byte {
	self.map_and_write_current_buffer()?;
	}
	}

This seems suboptimal from a performance point of view (given MappedWrite's primary use-case is to process the stdout of a process, which could be large in volume).

I happened to spot that the Rust stdlib actually has std::io::LineWriter which does something similar to what MappedWrite does.

Whilst we can't use LineWriter for our use-case (since it doesn't support a way to map the lines), it's possible we could take inspiration from its implementation:
https://github.com/rust-lang/rust/blob/master/library/std/src/io/buffered/linewriter.rs
https://github.com/rust-lang/rust/blob/master/library/std/src/io/buffered/linewritershim.rs