Building libCHIP.a via the Xcode build connector seems to involve codegen

[bzbarsky-apple]

There shouldn't be anything in libCHIP.a that depends on codegen, but what chip_xcode_build_connector.sh does right now (skipping some bits that I am pretty sure are irrelevant):

declare -a args=(
    'default_configs_cosmetic=[]' # suppress colorization
    'chip_crypto="boringssl"'
    'chip_build_tools=false'
    'chip_build_tests=false'
    'chip_enable_wifi=false'
    'chip_log_message_max_size=4096' # might as well allow nice long log messages
    'chip_disable_platform_kvs=true'
    "target_cpu=\"$target_cpu\""
    "target_defines=$target_defines"
    "target_cflags=[$target_cflags]"
    "mac_target_arch=\"$target_arch\""
    "mac_deployment_target=\"$LLVM_TARGET_TRIPLE_OS_VERSION$LLVM_TARGET_TRIPLE_SUFFIX\""
)
...
    args+=(
        'target_os="mac"'
    )
...
    gn --root="$CHIP_ROOT" gen --check out --args="${args[*]}"
    exec ninja -v -C out

ends up failing if ZAP is not installed. What's not clear to me is what about that command line is really saying "just build libCHIP.a", but the net effect is that there is just a libCHIP.a in out/lib (well, and libmbedtls.a for some reason and libnlunit-test.a) and no executables in out. So clearly the build is in fact just building things that should not depend on codegen.

We need to figure out two things:

1. Do these things in fact depend on codegen? If so, we need to fix that. But I suspect they do not in practice.
2. Is the build system just confused about whether they depend on codegen? If so, we need to fix that.

@mspang @andy31415

[bzbarsky-apple]

Looks like the dependency here may be through libChipController which uses CHIPClusters.h. Going to see if I can just remove that dependency.