refactor: generic unbounded cache

[janishorsts]

🛠️ Benefits, Issues Resolved & DX Impact

• Benefits: Replaced legacy interface{}-based caching with a strongly-typed generic synccache.Cache[K comparable, V any] implementation featuring zero-allocation cache hits (0 B/op), thread-safe single-flight request coalescing (inflight.go), and clean package scoping in internal/synccache.
• Issues Eliminated: Removed unsafe runtime type assertions and potential nil pointer dereferences across build context resolution callers (buildcontext, cachedmetaresolver, solvecache), and eliminated redundant concurrent load executions for identical in-flight keys.
• Developer Experience (DX) Impact: Compile-time type safety prevents type assertion bugs; simplified API surface replaces multi-layer context wrapping with direct shared context passing. Comprehensive test coverage added for concurrent loads, context cancellations, and race conditions.

⚡ Performance Impact

Metric	Before	After	Delta
Execution Time	105.4 ns/op	52.8 ns/op	-49.9%
Memory Allocations	16 B/op	0 B/op	-100.0%
Heap Mallocs	1 allocs/op	0 allocs/op	-100.0%

Note: Benchmarks measured on cache hits using Go 1.26 on darwin/arm64.

📦 Binary Size Impact

• Stripped Size (Before/After/Delta): 31.53 MB / 31.60 MB / +64.5 KB (+0.21%)

[gemini-code-assist]

Caution

The consumer version of Gemini Code Assist on GitHub has been sunset. All code review activity has officially ceased.

[github-actions]

➖ Are we earthbuild yet?

No change in "earthly" occurrences

📈 Overall Progress

Branch	Total Count
main	4748
This PR	4748
Difference	+0

---

Keep up the great work migrating from Earthly to Earthbuild! 🚀

💡 Tips for finding more occurrences

Run locally to see detailed breakdown:

./.github/scripts/count-earthly.sh

Note that the goal is not to reach 0.
There is anticipated to be at least some occurences of earthly in the source code due to backwards compatibility with config files and language constructs.

[gilescope]

@janishorsts have a look - it seemed there were a few holes (demonstrated by failing tests). I've patched them with claude's help and also switched from using sleeps to something more event based. (I try and avoid sleeps in tests wherever possible)

[gilescope]

Feel free to roll back changes you don't like / reimplement!

[janishorsts]

@janishorsts have a look - it seemed there were a few holes (demonstrated by failing tests). I've patched them with claude's help and also switched from using sleeps to something more event based. (I try and avoid sleeps in tests wherever possible)

@gilescope, nice. I had no idea golang introduced the "testing/synctest" for concurrency testing last year. I always read release notes.

Can you please have another look? I did a second round of changes, a significant refactoring. Initially, I was not keen on touching the metacontext package. But as it is used only by the cache implementation, I refactored it to be more idiomatic to Golang concepts.

[gilescope]

That's a big diff. Just seemed to be one issue around caching timeouts, please review my commit @janishorsts.