Support force: true in queued actions (batch close for no-reward networks)

[madumas]

Problem

When networks have no indexing rewards (e.g., fuse, xlayer-mainnet, xdai, blast-mainnet, monad, boba, mode), unallocate and reallocate actions queued via queueActions get stuck because the agent cannot generate a valid POI for these networks (see #1088).

The force: true flag works correctly on the direct closeAllocation mutation (with zero POI fallback from PR #949), but it is ignored when set on queued actions. This means the only way to close allocations on no-reward networks is through individual closeAllocation calls — there's no way to batch these via multicall.

Impact

• No batch/multicall for no-reward network closures — each must be an individual direct mutation call, one transaction per allocation
• Indexers managing allocations across many no-reward networks must work around the queue entirely, losing the benefits of batched execution
• The ActionInput type already accepts a force field, but the execution path doesn't honor it for queued actions

Proposed fix

When executing a queued unallocate or reallocate action that has force: true, the agent should use the zero-POI fallback path (same as the direct closeAllocation(force: true) mutation) instead of trying to query the EBO for a valid POI.

This would allow batching force-close actions alongside normal actions in a single multicall transaction.

Related

• Unable to close allocations on arbitrum-sepolia, sepolia chains without 0x000 and --force #1088 — EBO missing for no-reward mainnet networks
• PR common: allow closing of allocations on unsupported networks #949 — zero-POI fallback for unsupported networks (direct mutation only)