ci(ENG-21467): Make build-cpp even faster.

[markovejnovic]

What does this PR do?

Throws more jobs at cmake than cmake thinks it should use. Since cmake is oblivious to the fact that it's very likely going to hit a 100% hit rate, we need to throw more cores at it so that all the cache requests are made in parallel.

How did you verify your code works?

I ran CI with different job multipliers and compared the performance of each:

Job Multiplier	x64 mac	aarch64 mac	x64-asan	x64-musl	x64-baseline	win-x64-baseline	aarch64-musl	win-x64	x64	aarch64	x64-musl-baseline
1	7m21s	2m48s	FTC	5m31s	FTC	4m55s	4m58s	5m0s	FTC	FTC	5m41s
2	12m53s	2m46s	FTC	4m57s	FTC	4m45s	4m54s	5m1s	FTC	FTC	5m38s
4	12m48s	3m43s	FTC	5m30s	FTC	4m56s	4m46s	4m42s	FTC	FTC	5m14s
8	6m50s	2m47s	FTC	4m56s	FTC	4m46s	4m33s	4m43s	FTC	FTC	5m25s
16	8m51s	4m7s	FTC	5m3s	FTC	crash	4m51s	crash	FTC	FTC	5m2s
32	7m51s	2m50s	FTC	5m33s	FTC	crash	4m55s	crash	FTC	FTC	5m26s

FTC - Fail-to-cache https://linear.app/oven/issue/ENG-21502/x64-and-aarch64-linux-targets-failing-to-use-sccache
At high job multipliers, windows crashed due to high contention on the sccache lock-file in S3. I don't know why that happens.

Based on this plot, there is no appreciable difference in parallelizing the job, unfortunately

[linear]

ENG-21467 Throw more jobs at `build-cpp`

[robobun]

^{Updated 11:47 AM PT - Nov 11th, 2025}

❌ @markovejnovic, your commit 242917c has 5 failures in Build #31516 (All Failures):

• test/v8/v8.test.ts - 1 failing on 🐧 12 x64-asan
• test/cli/install/bun-install.test.ts - timeout on 🍎 13 x64
• test/cli/install/bun-install.test.ts - timeout on 🍎 14 aarch64
• test/cli/install/bun-install.test.ts - 1 failing on 🍎 13 aarch64
• test/js/node/worker_threads/worker_threads.test.ts - pid 64196 segmentation fault on 🍎 13 x64
• test/regression/issue/11297/11297.test.ts - 1 failing on 🪟 2019 x64-baseline
• test/cli/install/bun-install-lifecycle-scripts.test.ts - core dumped on 🐧 12 x64-asan

---

🧪   To try this PR locally:

bunx bun-pr 24452

That installs a local version of the PR into your bun-24452 executable, so you can run:

bun-24452 --bun

[markovejnovic]

As per the discussion in the PR, there is no appreciable difference with parallelization so I think it safe to reject this PR.