fix(aptos): publish the version lower bound on the slot axis

[msizov]

In production, aggregator_upstream_providers_lower_bounds{chain="aptos"} has only type="lower_block" — no version bound at all. The live upstream description confirms it:

lower_bounds:{k:-0  b:1.049154233e+09  key:LOWER_BOUND_BLOCK}
status:{current_height:1053521899  slot:7256690171  availability:AVAIL_OK}

That left dproxy rejecting every version-addressed aptos request (/v1/blocks/by_version, /v1/transactions/by_version) for having no bound data — 100% failure for a customer running Chainlink.

Cause

internal/protocol/upstream_state_events.go:35

if head := state.HeadData.Height; l.Data.Type != SlotBound && head > 0 && l.Data.Bound > int64(head) {
    return true // "Same" -> event skipped, bound never applied
}

The detector published oldest_ledger_version (~7.23e9) as StateBound, but HeadData.Height for aptos is the block height (~1.05e9). Ledger versions run ~7x ahead, so every publish was discarded as a bound above the head.

Fix

Publish it as SlotBound, the one type exempt from that comparison, and the axis the head already carries — GetLatestBlock passes the ledger version as the block's Slot (aptos_chain_specific.go). This matches how solana already splits its two bounds. BlockBound is unchanged.

Verification

Patched nodecore against a drpc aptos fullnode (fornex-eu-bcn-20), reading the chain-status stream a gRPC client sees:

chain=CHAIN_APTOS__MAINNET LOWER BOUNDS:
  type=LOWER_BOUND_SLOT  value=7230467940
  type=LOWER_BOUND_BLOCK value=1049154233
chain=CHAIN_APTOS__MAINNET HEAD height=1053709036 slot=7258271398

Both match the node's own oldest_ledger_version / oldest_block_height, and the slot bound now survives even though 7.23e9 ≫ the head height.

Notes

• Ship together with drpcorg/dproxy#2958, which routes aptos version requests on the slot axis. Landing this first would break archive-backed aptos upstreams: they previously published StateBound=1, which is below the head and so survived the guard, and would publish nothing until the dproxy side lands. No such upstream exists today, but don't rely on that.
• This removes the last sanity check on the bound — SlotBound is exempt from the above-head comparison and there is no slot-axis equivalent, while Same rejects decreases unless the value is exactly 1. A bogus oversized oldest_ledger_version would now stick until restart. Worth a HeadData.Slot guard as a follow-up.