build/install_cmake/ninja: use project instead of script

This is more compact and robust

Use CMake scripts to install

The Python cmake/ninja install scripts had duplicated, outdated logic compared to the CMake scripts which are more widely uses, so we removed the cmake_setup and ninja_setup from Python.

Instead, recommend the "install_cmake.cmake" and "install_ninja.cmake" scripts that are now tested across OS in CI

use JSON for versions and URLS

also numerous refactoring and quality improvments

cleanup deprecated functions

v1.9.0

parameter template

default cmake 3.20

v1.8.4

v1.8.4

default CPU count=1 to not crash make

Make defaults to unlimited large number of builds when -j is invoked. So we made a "safe" default of 1 when CPU count or RAM cannot be determined

better autocpu count

• address github CI path issue
• cmake_compile: more accurate CPU count, including with Make, better/faster builds without crashing

compile: more parallel, -workdir option

• allow cmake_compile to build with all CPU if system has enough RAM
• -workdir option allows using own development branch of CMake

use entry_points, modern cmake syntax

• use entry_points since that's much more convenient for day-to-day use than a script laying around somewhere
• cmake_compile: use tempdir and modern CMake command syntax

generalize cmake_setup.py to allow multiple windows installs

use .zip file for Windows binaries to allow multiple Windows CMake installs, which is not possible with the .msi installer. The .zip method is much faster and automated vs. .msi.

Deduplicated and made cmake_setup more robust, for example checking that requested version is available rather than giving HTTPerror