feat(ledger): map Unichain, so its deployments are no longer skipped

[venugopalanvip]

The gap

ledger_network_id returns None for chain 130, so convert_erc7730_v2_input_to_calldata drops every deployment on it:

if ledger_network_id(deployment.chainId) is None:
    out.warning(f"Chain id {deployment.chainId} is not known, skipping it")
    continue

No device descriptor is produced for that chain, and the only signal is a warning on a converter run that still exits successfully.

Ledger does support Unichain. @ledgerhq/cryptoassets@13.56.0 carries both:

unichain          { type: "CryptoCurrency", family: "evm", name: "Unichain",         chainId: 130  }
unichain_sepolia  { type: "CryptoCurrency", family: "evm", name: "Unichain Sepolia", chainId: 1301 }

Both ids are taken verbatim from that package rather than guessed. As a check on the source: of the 123 chain ids this file already maps, 61 also appear in the package, and all 61 agree exactly — so it is the right list, and the ids are in the right form.

What it recovers

Seven descriptors in the clear-signing registry declare a deployment on one of the two and lose it today:

descriptor	recovered
registry/ekubo/calldata-MEVCaptureRouter.json	+2 of 18
registry/kyberswap/calldata-MetaAggregationRouterV2.json	+1 of 18
registry/layerswap/calldata-LayerswapDepository.json	+1 of 56
registry/lido/calldata-stETH-L2.json	+1 of 2
registry/lido/calldata-wstETH-L2.json	+1 of 6
registry/okx/calldata-OkxDexRouterV1.0.7-multi-commission.json	+1 of 25
registry/okx/calldata-OkxDexRouterV1.0.8-suffix-compat.json	+1 of 25

lido/calldata-stETH-L2.json is the sharpest: it declares two deployments, and one of them is Unichain.

The wider measurement, and why this PR stops here

I swept every calldata descriptor in the registry to find how much this costs in total:

	count
distinct chain ids used by registry calldata descriptors	95
known to ledger_network_id	48
unknown, deployments silently dropped	47
deployments affected	87
descriptors affected	16

So roughly half the chains the registry targets produce no device output. But of those 47, only these two are in the cryptoassets package. The other 45 — Plasma (9745, 12 deployments), Soneium (1868), TAC (239), MegaETH (4326), Manta Pacific (169), opBNB, Fraxtal, Ronin, Taiko, Zora and the rest — have no Ledger currency id at all, so there is nothing correct to map them to. Dropping them is the right behaviour, not a bug, and inventing ids for them would be worse than the warning.

That is why this PR adds two lines rather than forty-seven. If Ledger adds currencies for any of those chains, the same one-line-per-chain change applies; happy to file the full list as an issue if it is useful for tracking.

Tests

tests/common/test_ledger_network_id.py is new — the file had none. It pins the two new ids, an existing one either side, the None for an unsupported chain, and that the match arms stay in ascending chain-id order, which is what makes the insertion point for the next chain unambiguous.