feat(eval): implement holiday support for NETWORKDAYS and WORKDAY

[TyrGo]

The holidays parameter (third argument) was accepted but silently ignored in both NETWORKDAYS and WORKDAY, producing incorrect results when callers supplied holiday dates.

Changes:

• Add collect_holidays() helper that parses the holidays argument from a cell range, inline array constant, or single scalar value into a HashSet for O(1) lookups during day iteration.
• Add literal_to_serial() helper for coercing LiteralValue variants to date serials within flattened holiday arrays.
• Wire both NETWORKDAYS and WORKDAY eval bodies to use collect_holidays(), replacing the empty Vec placeholders.
• Update docstrings and inline examples to reflect correct behavior.
• Add 10 unit tests covering: no holidays, single scalar holiday, array holidays, holiday on weekend (no effect), reversed date range, negative workday offset, holiday on start date, and native Date literal holidays.
• Add 6 JSON formula integration tests for range and scalar holiday inputs via the formula test runner.

[Copilot]

Pull request overview

Adds correct holiday exclusion support to the date/workday builtins so callers supplying a holidays argument get accurate results (matching typical spreadsheet behavior), and extends coverage via unit + JSON integration tests.

Changes:

• Add holiday parsing/coercion helpers (collect_holidays, literal_to_serial) and use a HashSet<NaiveDate> for O(1) holiday lookups.
• Wire NETWORKDAYS and WORKDAY to exclude/skip holidays in addition to weekends.
• Update docs/examples and add Rust unit tests + JSON formula runner tests for holiday scenarios.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 4 comments.

File	Description
crates/formualizer-eval/src/builtins/datetime/weekday_workday.rs	Implements holiday collection/coercion and uses it in NETWORKDAYS/WORKDAY, plus adds unit tests and doc updates.
tests/formula_tests/date_time.json	Adds formula-runner integration tests validating holiday handling for scalar and range inputs.

---

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

[Copilot]

collect_holidays(args, 2)? will propagate coercion/serial conversion errors as a returned Err from eval, which the formula test runner treats as an evaluation failure rather than an in-cell #VALUE!/#NUM! result. Since the holidays argument uses ArgSchema::any() (no coercion/validation), consider catching Err(e) here and returning Ok(CalcValue::Scalar(LiteralValue::Error(e))) so invalid holiday inputs and holiday-range errors behave like regular Excel errors instead of aborting evaluation.

Suggested change

	let holidays = collect_holidays(args, 2)?;
	let holidays = match collect_holidays(args, 2) {
	Ok(h) => h,
	Err(e) => {
	return Ok(CalcValue::Scalar(LiteralValue::Error(e)));
	}
	};

[Copilot]

Same issue as NETWORKDAYS: collect_holidays(args, 2)? can cause WORKDAY to return an Err(ExcelError) for bad holiday inputs (text that can’t be parsed, out-of-range serials, errors inside a holiday range), which surfaces as an evaluation failure instead of an in-cell error value. Handle Err(e) by returning Ok(CalcValue::Scalar(LiteralValue::Error(e))) at this callsite.

Suggested change

	let holidays = collect_holidays(args, 2)?;
	let holidays = match collect_holidays(args, 2) {
	Ok(h) => h,
	Err(e) => {
	return Ok(CalcValue::Scalar(LiteralValue::Error(e)));
	}
	};

[TyrGo]

I considered this, but every other error-returning call in this file uses ? propagation (coerce_to_serial(&args[0])?, coerce_to_int(&args[1])?, etc.). Wrapping just collect_holidays in a match would be inconsistent with the established pattern. If error wrapping should be preferred over ? propagation, that seems like a codebase-wide decision rather than something to change for one callsite.

[TyrGo]

Covered by recent merge. Closed.