feat: consolidate probing; rename @lde/sparql-monitor → @lde/distribution-monitor

[ddeboer]

Summary

Makes @lde/distribution-probe the single source of truth for probing (including SPARQL endpoints), and renames @lde/sparql-monitor to @lde/distribution-monitor to reflect its broader scope – it now monitors any DCAT distribution, SPARQL endpoint or data dump.

This is a breaking-change PR. Pre-release status per CLAUDE.md, so no shims.

Part A – Extend @lde/distribution-probe

• probe() now takes ProbeOptions (timeoutMs, headers, sparqlQuery). Breaking: the old probe(distribution, 5000) number form is gone; pass { timeoutMs: 5000 }. @lde/pipeline’s one call site is migrated in this PR.
• SparqlProbeResult, DataDumpProbeResult, and NetworkError all carry a new responseTimeMs: number field, measured from just before the fetch to just after the response is fully consumed (or the error is thrown). Breaking: the result-class constructors take this as a required parameter; SparqlProbeResult additionally takes acceptedContentType. Pipeline test fixtures that construct these directly are updated.
• URL-embedded Basic auth (user:pass@host) is extracted into an Authorization: Basic … header. Applied to both SPARQL and data-dump probes; caller-supplied Authorization wins on conflict.
• SPARQL probing dispatches on query type (ASK / SELECT / CONSTRUCT / DESCRIBE). The Accept header and response validation are query-type-aware (JSON boolean for ASK, JSON results for SELECT, non-empty RDF body for CONSTRUCT / DESCRIBE). Query-type detection uses an inline regex with #-comment stripping.
• Design note on fetch-sparql-endpoint: the task suggested delegating to SparqlEndpointFetcher.fetchAsk/fetchBindings/fetchTriples for query dispatch. I kept raw fetch instead so SparqlProbeResult retains the underlying Response (statusCode, statusText, contentType, lastModified). SparqlEndpointFetcher throws on HTTP errors and hides the response, which is awkward when the probe’s job is to describe failures. The inline regex is the task’s explicitly-allowed fallback.

Part B – Rename @lde/sparql-monitor → @lde/distribution-monitor

• New config shape: { identifier, distribution: { accessUrl, mediaType?, conformsTo? }, sparqlQuery? }. Any DCAT distribution is now supported; sparqlQuery only applies to SPARQL endpoints.
• SparqlMonitor class removed; MonitorService now calls probe() directly and maps the returned ProbeResultType onto the existing CheckResult shape (failing probes become success: false with failureReason/joined warnings/HTTP-status as the error message).
• PostgreSQL schema is unchanged – CheckResult’s shape is already minimal enough to cover both cases. Adding a probe_type column is noted as a possible follow-up.
• CLI bin is distribution-monitor. Breaking: no sparql-monitor alias.
• MonitorServiceOptions gains timeoutMs, headers, and an injectable probe (for tests).

@lde/sparql-monitor deprecation

packages/sparql-monitor/ is deleted from the workspace. No final re-export shim is published. @lde/sparql-monitor on npm stops receiving updates; consumers must migrate to @lde/distribution-monitor directly.

Post-merge action for the maintainer (outside this PR): run npm deprecate @lde/sparql-monitor "Renamed to @lde/distribution-monitor" so npm install of the old name surfaces a warning, and enable npm Trusted Publisher on @lde/distribution-monitor after its first publish.

Out of scope

• Structured probe-failure taxonomy (DnsFailure / ConnectionTimeout / TlsError): defer until a second consumer requests it.
• Changes to probeResultsToQuads output in @lde/pipeline/src/distribution/report.ts.
• Extending the PostgreSQL schema beyond what the rename needed.