Build Roman on Apple Silicon

Updates for building Roman development system on Apple Silicon macs.

This PR is related to the following issue but does not directly address the issues presented there:

https://github.com/spacetelescope/science-platform-images/issues/32

Adds --platform to the Docker build command.
Starts on conversion from DockerHub to quay.io for builds using pre-built jupyter by adding REGISTRY and consolidating DOCKER_BUILD_ARGS.
Switches to formal release requirements to development requirements.

In theory I used this Colima setup:

colima start \
--profile intel \
--arch x86_64 \
--cpu 10 \
--disk 200 \
--memory 32 \
--vm-type vz \
--vz-rosetta \
--verbose

Also POSSIBLY IMPORTANT as a work around to failures fetching manifests from DockerHub, I performed my successful laptop build OFF of our VPN.

My setup-env was:

# Clone and customize this for your deployment
#
# NOTE:  must be sourced from jupyterhub-deploy root directory
#

# ----------------- basic inputs,  must set for deployment --------------

export DEPLOYMENT_NAME=roman        # roman, jwebbinar, tike
export ENVIRONMENT=sandbox          # sandbox, dev, test, prod, int
export CAL_VERSION=none             # e.g. roman or jwst x.y.z CAL release, latest, or none

export USE_FROZEN=0                  # use 0 for loosely pinned package versions,
                                     # 1 for tagged production build, 2 for chilly

export FREEZE_CHILL=1                # capture new frozen and chilly requirements as
                                     # modified files in this git repo

export PLATFORM="linux/amd64"        # os / instruction-set for built image: linux/amd64 or linux/aarch64

export REGISTRY=""
# export REGISTRY="quay.io/repository/"    # empty for DockerHub,  but jupyter/docker-stacks is now at quay.io/repository

export OWNER="spacetelescope"          # jupyter or spacetelescope base image repo
# export OWNER="jupyter"          # jupyter or spacetelescope base image repo
# export OWNER="${REGISTRY}/jupyter"          # jupyter or spacetelescope base image repo
                                     # jupyter will be pulled from DockerHub
                                     # spacetelescope will be built from source code

export BASE_IMAGE=${OWNER}/scipy-notebook   # from jupyter/docker-stacks

export NOTEBOOK_TAG=latest
export NOTEBOOK_ID=notebook-${DEPLOYMENT_NAME}:${NOTEBOOK_TAG}  # abstract name of mission-specific notebook image

# ----------------- vvvv less frequently changed vvvv -------------------------------

source infrequent-env   # for image development

This runs fairly well but two import tests fail due to numpy v1/v2 version build issues.

spacetelescope / science-platform-images

Build Roman on Apple Silicon #33