feat: indexed aggregates behind --indexed-aggregates

[SgtPooki]

What changed

Adds an opt-in indexed-aggregate mode: plan --indexed-aggregates packs sub-pieces into FRC-0058 data segment aggregates, redirect-serve streams the assembled aggregate at /piece/{root} (each sub-piece hash-verified on the way out), and pdp-submit pulls that one stream and adds the root as its own single sub-piece — the 1:1 shape Curio's v0 add requires. The layout is built in pure JS (ipfs2foc-core/indexed-aggregate) from the @web3-storage/data-segment dep, with offsets and index serialization matched against go-data-segment source. Default (bare) aggregation is unchanged.

This targets the provider-side aggregate indexing being ported from Curio's mk20 pipeline into the PDPv0 path; keep on this branch until that lands and a calibration round-trip passes.

How to verify

pnpm -r typecheck && pnpm biome check . && pnpm -r test — the key tests recompute the piece commitment over the assembled bytes (unit and over real HTTP) and require it to equal the aggregate root, the same check the provider runs on pull.

Notes / risks

New indexed column on aggregates (pre-release, no migration). Indexed aggregates need --source-base; the stateless relay cannot assemble the stream.

[SgtPooki]

closing this because its using an old optimization attempt that is a little more complex than we want to offer to self-serve users. However, for those more technically minded and wanting to optimize certain avenues, this may be a reasonable approach for you.