Previously, BootZipCopyAction would put the next entry to the stream
and then, in the case of a stored entry, configure its size, CRC32
etc. This had the benefit of being able to copy the entry into the
zip once, capturing its bytes for the calculation of the CRC32 as it
was copied. Unfortunately, while this produced zip files that could
be read by the JVM, other zip tools failed. For example, Go's zip
support that's used by CloudFoundry could not unzip the archive.

This commit updates BootZipCopy action to completely configure the
entry before putting it to the stream. This has the downside of
copying the file twice (once for the CRC32 and once to actually write
it to the zip stream) but this appears to be unavoidable as we have to
produce archives that can be unzipped without problems on all
platforms.

Closes gh-8816