chore(deps): update synapse-core to 0.8.1 and synapse-sdk to 1.2.1

[nijoe1]

Summary

• Update @filoz/synapse-core 0.7.0 → 0.8.1 and @filoz/synapse-sdk 1.1.0 → 1.2.1.
• Adopt the one breaking change (core 0.8.0, Add pagination and expanded data set, piece, provider, and payment queries FilOzone/synapse-sdk#913/#914): list queries are cursor-paginated ({items, nextCursor} pages, default 100/page), and activePieceCount is removed in favor of a bounded hasActivePieces flag.
• dataset list and wallet costs walk every page via the SDK's paginate() generator, preserving the old return-everything behavior for wallets with more than 100 datasets.
• piece list and dataset details switch from --offset to --cursor; output hasMore/nextOffset become an optional nextCursor (absent = last page). The "fetch all N pieces" CTA goes away with the total count; the next-page CTA carries the exact --cursor command instead. dataset details reports hasActivePieces.
• utils.ts uses the renamed FilecoinChain type export.
• SKILL.md pagination docs and examples updated; the schema-gating test enforces them.

Compatibility note

The JSON output of piece list and dataset details changes shape: hasMore/nextOffset are replaced by an optional nextCursor string, and the dataset object reports hasActivePieces: boolean instead of activePieceCount. Cursors are opaque continuation values — pass nextCursor back as --cursor until it stops appearing.

Upstream notes

• getPiecesWithMetadata is deprecated upstream (removal planned in feat!: remove piece metadata getters FilOzone/synapse-sdk#936; replacement is getPieces plus PieceAdded events or an indexer). It still works in 0.8.1, so this PR keeps it and defers the rewrite until removal actually ships.
• The bump also picks up the filecoin-services v1.3.1 mainnet addresses bundled in getChain(), hardened provider health checks, and read-only calls that omit the sender.
• claimTokens is now strictly typed as the two-asset tuple with faucetInfo — matching what fix(wallet): report per-asset faucet outcomes #35 maps, so no rework needed there.

Validation

• bun run test: 203 passed, 496 assertions (includes the schema-gated skill examples)
• bunx biome check src tests
• bunx tsc --noEmit
• bun run build

[snissn]

Reviewed exact head 0280a260b69067707facfad117510163b2633efd. The dependency migration and one-page mappings look sound, and local validation passed: bun run test (203 tests / 496 assertions), Biome, TypeScript, and build.

Changes required before merge:

1. Preserve the opaque cursor without passing it through JavaScript number; the inline comment has the concrete failure and scope.
2. Add a real two-page regression test for both dataset list and wallet costs. Their current tests return only one page, so the headline behavior of retaining all datasets beyond the first 100 is not exercised. Mock page 1 with nextCursor, return page 2 for that exact cursor, and assert both cursor progression and accumulated results/cost inputs.

GitHub CI is green on this head and there were no pre-existing review threads. No merge performed.

[nijoe1]

Both requested changes are addressed on head 2c39b27:

1. Cursor precision — --cursor is a validated decimal string in piece list and dataset details, converted with BigInt only at the SDK boundary; the next-page CTA and nextCursor output carry the string unchanged. Regression tests cover the exact round-trip at 2^53 + 1 for both commands. Thread resolved.
2. Two-page coverage — dataset list and wallet costs now have regression tests where page 1 returns a nextCursor, page 2 answers that exact cursor, and the assertions check cursor progression plus the accumulated datasets / cost inputs.

Validation: bun run test 207 passed / 509 assertions, bunx biome check src tests, bunx tsc --noEmit, bun run build.

[snissn]

Follow-up reviewed exact head 2c39b27285df979e1867c78b6ad818f4383f8f5f.

Both requested changes are resolved:

1. piece list and dataset details now validate cursors as decimal strings, convert to BigInt only at the SDK boundary, and preserve exact CTA/output values. Both commands have regression coverage above 2^53.
2. dataset list and wallet costs now have real two-page tests asserting the exact continuation cursor and accumulated dataset/cost inputs.

Validation on this exact head: bun run test (207 passed / 509 assertions), Biome, tsc --noEmit, and build. GitHub CI is green, the prior inline thread is resolved, and the head merges cleanly with current main.

Approved. No merge performed.