Move module loading out of js::core::Context, so it can be customised

[eddyashton]

Goal is that other apps can use the js::core::Context, but read JS code from their own tables. Written to be extensible so they could also load some from hardcoded values, or modify/combine values in the KV, or...

[achamayou]

@eddyashton lgtm but deserves a nightly run I think.

[eddyashton]

Note for posterity: A previous version of this PR introduced a major memory leak*, that we almost missed as we currently have no visibility into perf regressions. @achamayou stated this explicitly - we're currently making large changes to the JS interpreter but flying blind with regard to the performance impact. We only caught this one pre-merge by chance, because of a single (Daily build-gated) stress test which ballooned the memory so far it crashed a node. Restoring perf visibility is a high priority.

* I moved the cache of previously-loaded modules out of the Context, and into one of its Loaders. This meant we weren't getting caching benefits across invocations, since we installed new Loaders each time, so a long-lived interpreter (re-fetched from the interpreter cache) was continually re-processing the same modules, and linearly growing its memory use.

[github-actions]

Bencher

Report	Mon, June 3, 2024 at 09:19:58 UTC
Project	CCF
Branch	6199/merge
Testbed	gha-virtual-ccf-sub

⚠️ WARNING: The following Measure does not have a Threshold. Without a Threshold, no Alerts will ever be generated!

• Latency (latency)

Click here to create a new Threshold
For more information, see the Threshold documentation.
To only post results if a Threshold exists, set the --ci-only-thresholds CLI flag.

Click to view all benchmark results

Benchmark	Latency	Latency Results nanoseconds (ns)
commit_latency_ratio	➖ (view plot)	0.74
historical_queries	➖ (view plot)	61,028.59
pi_basic_js_virtual	➖ (view plot)	4,360.70
pi_basic_virtual	➖ (view plot)	57,701.50

Bencher - Continuous Benchmarking
View Public Perf Page
Docs | Repo | Chat | Help