[dev/build] use more performant copy implementation

[spalger]

Was looking through the build for things that could be sped up, one thing that stood out is the Creating platform-specific archive source directories step. It currently takes about 2.5 minutes on CI and I figured it was something similar to the deep delete slowness we fixed with #24692, so I gave a custom implementation a shot and right away saw massive speed improvements. The current implementation copies the archive source per-platform in 6 seconds on CI 👍

00:10:39.912  info [kibana-oss] Creating platform-specific archive source directories
00:10:45.963    │ debg Generic build source copied into linux specific build directory
00:10:46.346    │ debg Node.js copied into linux specific build directory
00:10:46.346    │ succ ✓ 6 sec

[elasticmachine]

Pinging @elastic/kibana-operations

[elasticmachine]

💚 Build Succeeded

• continuous-integration/kibana-ci/pull-request

[elasticmachine]

💚 Build Succeeded

• continuous-integration/kibana-ci/pull-request

[mistic]

LGTM! I tested this locally and it works quite well! I'm just wondering if we have more of those things spread out in other places across our build! We are also using the old copy_all into the copy_source_task and in the optimize_task

[spalger]

We could use this in more places, but I'm not trying to make a copy function that solves every need (though it might already solve more than I'm trying to fix). This is purely focused on performance.

[spalger]

6.x: ac90d3b