Better build orchestration

The underlying build system of rbwasm build is quite messy.

1. Each build step has to have its ad-hoc up-to-date check like followings:
   https://github.com/ruby/ruby.wasm/blob/a5d06a01873130fda0f07f4401a3b018c6f1d753/lib/ruby_wasm/build/product/baseruby.rb#L26 https://github.com/ruby/ruby.wasm/blob/a5d06a01873130fda0f07f4401a3b018c6f1d753/lib/ruby_wasm/build/product/crossruby.rb#L171
2. Uses of build product and cache key computation are placed far, so we often forget to update the cache key and get a wrong cache restored.
   https://github.com/ruby/ruby.wasm/blob/a5d06a01873130fda0f07f4401a3b018c6f1d753/lib/ruby_wasm/build.rb#L78
3. The current cache mechanism cannot restore two or more artifacts produced within a single rake task job. This usually happens for PIC & non-PIC builds invoked by rake npm:ruby-head-wasm-wasi:build
4. Our cache mechanism does not allow partial cache-hit, so even if we change only crossruby source revision, we have to rebuild openssl from scratch.
5. Our cache mechanism is not sensitive to recipe changes. For instance, even if we change CFLAGS value, crossruby product will not be rebuilt.

首先阅读 lib/ruby_wasm/build.rb，以及 lib/ruby_wasm/build/product/baseruby.rb 和 crossruby.rb 中的产品实现。跟踪入口 rake npm:ruby-head-wasm-wasi:build 及其 PIC 和非 PIC 构件。完成标准是构建缓存能够处理多个构件和部分命中，并能处理源代码修订版本和配方的变更，而不会恢复错误的结果。