Improve scheduler/oracle test coverage and stabilize CI workflow runtime/toolchain

[Copilot]

This PR addresses low Go test coverage in core runtime packages and resolves CI instability in the existing pipeline. It adds targeted unit coverage for scheduler/oracle logic and updates workflow runtime/toolchain configuration to match current project and ecosystem requirements.

• Coverage expansion: scheduler core paths

  • Added focused unit tests for provider lifecycle and invariants:
    • register/get/getActive semantics
    • active-provider counting
    • deregistration behavior
    • reputation clamping bounds
    • HTTP handler validation and response behavior
  • Added scheduler-level tests for:
    • provider eligibility and assignment behavior
    • no-eligible-provider path (pending/unassigned behavior)
    • job retrieval handler behavior
    • metrics output correctness

• Coverage expansion: scoring and oracle behavior

  • Added scoring tests to validate:
    • weighted scoring behavior
    • oracle-price override impact
    • clamping/normalization expectations
  • Added oracle tests for:
    • set/get/getAll semantics
    • not-found handling in price lookup endpoint
    • pricing feed/jitter expectations

• Build/CI correctness fixes

  • Implemented missing ProviderManager.GetCount() used by scheduler metrics (removes scheduler package build failure in test job).
  • Updated GitHub Actions workflow to current action/runtime baselines:
    • actions/checkout → v4
    • actions/setup-go → v5 with go-version-file: go.mod
    • actions/setup-node → v4 with Node 22.10.0 (Hardhat-compatible)

Example of the scheduler build fix:

func (pm *ProviderManager) GetCount() int {
	pm.mu.RLock()
	defer pm.mu.RUnlock()

	count := 0
	for _, p := range pm.providers {
		if p.Active {
			count++
		}
	}
	return count
}

[Copilot]

Pull request overview

This PR improves reliability and confidence in the scheduler/oracle services by adding targeted Go unit tests and aligning CI tooling/runtime versions with the repo’s current Go/Node requirements.

Changes:

• Added new unit tests covering scheduler scoring/assignment paths, provider manager semantics/handlers, and oracle handlers/pricing helpers.
• Implemented ProviderManager.GetCount() to support scheduler metrics and fix a build failure.
• Updated GitHub Actions workflow to newer actions/* versions and to use Go from go.mod plus a pinned Node version for Hardhat.

Reviewed changes

Copilot reviewed 7 out of 7 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
scheduler/scoring_test.go	Adds scoring tests for oracle overrides and clamping behavior.
scheduler/scheduler_test.go	Adds tests for job submission, handlers, and metrics output.
scheduler/provider_manager_test.go	Adds tests for provider lifecycle, handler validation, and JSON responses.
scheduler/provider_manager.go	Adds GetCount() for active provider counting (used by metrics).
oracle/pricing_test.go	Adds tests for pricing feed contents and jitter bounds.
oracle/oracle_test.go	Adds tests for oracle set/get semantics and HTTP endpoints.
.github/workflows/ci.yml	Updates action versions and toolchain configuration for Go/Node.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

This test doesn't validate the "array not null" behavior: unmarshalling null into a slice results in nil with len==0, so the test would pass even if the handler encoded null. To assert the response is an empty array, compare the raw body (trimmed) against [], or unmarshal into *[]PriceEntry and ensure the pointer is non-nil.

[Copilot]

This test submits a job while the scheduler is configured with oracleURL set to http://127.0.0.1:0. During assignment, fetchOraclePrice will attempt a real HTTP request to that address (connection error -> fallback), introducing unnecessary network dependency/slowness. Consider using an httptest.Server (even one returning 404/empty) for oracleURL so the test stays fully hermetic.

[Copilot]

ScoreProviders returns results in the same order as the input slice and does not rank/sort them. This test currently asserts results[0] is p1, which will be true even if p1 scores lower. Sort results by Score before asserting the top provider, or compare the computed scores directly (e.g., results[0].Score > results[1].Score after locating entries by ID).

Suggested change

	if results[0].Provider.ID != "p1" {
	t.Fatalf("expected p1 to score higher with oracle price override, got %s", results[0].Provider.ID)
	}
	var p1Result, p2Result *ScoredProvider
	for i := range results {
	switch results[i].Provider.ID {
	case "p1":
	p1Result = &results[i]
	case "p2":
	p2Result = &results[i]
	}
	}
	if p1Result == nil || p2Result == nil {
	t.Fatalf("expected results for both p1 and p2")
	}
	if p1Result.Score <= p2Result.Score {
	t.Fatalf("expected p1 to score higher with oracle price override, got p1=%f p2=%f", p1Result.Score, p2Result.Score)
	}

[Copilot]

This test doesn't actually distinguish between null and [] in the JSON response: unmarshalling null into a slice yields a nil slice with len==0, so the assertion still passes. To verify the handler returns an empty array, inspect the raw response bytes (after trimming whitespace) and assert it equals [], or unmarshal into *[]Provider and assert the pointer is non-nil.

Suggested change

	var providers []Provider
	if err := json.Unmarshal(rec.Body.Bytes(), &providers); err != nil {
	t.Fatalf("decode response: %v", err)
	}
	if len(providers) != 0 {
	t.Fatalf("expected empty provider list, got %d", len(providers))
	}
	body := strings.TrimSpace(rec.Body.String())
	if body != "[]" {
	t.Fatalf("expected empty JSON array [], got %q", body)
	}