Fix b with zsh.

zsh doesn't do string splitting like bash, so it passes the
post-processed bazel args to bazel as a single arg. So this CL adds a
conditional to do the splitting correctly for zsh with `setopt
shwordsplit`.

Reference: https://zsh.sourceforge.io/FAQ/zshfaq03.html

For bash, this uses array[@] to split the list into separate
shell words.

Test: zsh; source build/envsetup.sh && b test
//system/logging/logd:logd-unit-tests --config=linux_x86_64
Test: bash; source build/envsetup.sh && b test
//system/logging/logd:logd-unit-tests --config=linux_x86_64

Change-Id: I4e19a062b7f7e119b1612a8ce5c801878378cc69