perf(build): compiled the kernel and each userland library and app in parallel

[FlareCoding]

Summary

• The kernel and userland builds ran one compile at a time, so a clean userland build used one core and took about 200 seconds on CI's 4-core runners.
• The kernel and userland builds run with -j$(NPROC), and .NOTPARALLEL keeps userland's own steps in order: the SDK, then the libraries, then the apps, then package staging. Only the compiles inside each library and app run in parallel, so the build output and progress lines read as before.
• A parallel build produces the same binaries as a serial one, apart from vim, which embeds its build time.

---

Note

Low Risk
Makefile-only build parallelism with ordered userland phases preserved; no runtime or security impact.

Overview
Speeds up clean builds by passing -j$(NPROC) into the top-level kernel and userland sub-makes so compilation can use all available cores (e.g. CI’s 4-core runners instead of a single-threaded ~200s userland build).

In userland/Makefile, .NOTPARALLEL keeps the existing pipeline order—SDK, then libraries, then apps, then package staging—so progress output stays readable; parallelism is limited to compiles within each library/app (and the kernel tree), not across those phases.

Risk: build orchestration only; output is intended to match serial builds aside from vim embedding its build time.

^{Reviewed by Cursor Bugbot for commit c6716d8. Bugbot is set up for automated code reviews on this repo. Configure here.}