Fix sweep nonce non-determinism and disabled TLS hostname verification

[RobGanon]

ElectrumSVP: fix two sweep-path security issues

Attach electrumsvp-fixes.patch (or paste the diff below) to an Issue — or open it as a PR against TruthMachine/ElectrumSVP.

Problem 1 — non-deterministic ECDSA nonce in sweep signing

electrumsv/gui/qt/sweep_helper.py signs sweep txs with
ecdsa.SigningKey.sign_digest(...), which — because no k is supplied —
falls back to os.urandom for the nonce (the ecdsa docs themselves say to
use sign_digest_deterministic instead). With a weak/broken RNG or nonce
reuse, an attacker can recover the swept private key. Everywhere else the
wallet uses RFC6979 (bitcoinx PrivateKey.sign → libsecp256k1); the sweep
path is the only exception.

Fix: sign with an RFC6979 deterministic nonce:

-        der_sig = sk.sign_digest(sighash, sigencode=ecdsa.util.sigencode_der_canonize)
+        der_sig = sk.sign_digest_deterministic(
+            sighash, hashfunc=hashlib.sha256, sigencode=ecdsa.util.sigencode_der_canonize)

hashlib is already imported in this module. Output format is unchanged
(low-S canonical DER + 0x41 sighash suffix).

Problem 2 — TLS hostname verification disabled

sweep_helper.create_ssl_context() sets ctx.check_hostname = False for the
direct ElectrumX socket fallback, disabling hostname verification (normal
wallet connections via aiorpcx don't do this). Fix: remove that line.

             ctx.load_verify_locations(cafile=pem_path)
-            ctx.check_hostname = False

Verification performed

• Confirmed old path emits different signatures on repeated signings.
• Confirmed new path is deterministic (RFC6979), same low-S canonical DER
  structure, and the signature verifies against the public key.
• Confirmed the SSL context keeps check_hostname=True / CERT_REQUIRED.
• py_compile passes; the only changes are in sweep_helper.py (2+/2-).

How to apply

git apply electrumsvp-fixes.patch

or

patch -p1 < electrumsvp-fixes.patch

[TruthMachine]

I investigated both changes separately. The deterministic-signature issue is already resolved in my development branch, where sweep signing uses bitcoinx/libsecp256k1 rather than the ecdsa signing path in this PR.

Regarding TLS hostname verification, I agree that disabling hostname verification weakens TLS security. However, ElectrumSVP currently uses server-specific self-signed certificates, and I verified that at least sv2.satoshi.io currently presents a self-signed certificate for CN=electrumr with no SAN for sv2.satoshi.io. With that certificate explicitly trusted, hostname verification succeeds only when check_hostname=False; enabling it produces a hostname-mismatch failure. Simply removing check_hostname=False would therefore break some existing configured servers.

This is not ideal, but in the specific case of the sweep operation, a MITM would not gain access to the private keys. They could potentially provide incomplete or incorrect UTXO data, which could cause a sweep to fail or result in fewer UTXOs being included, but they would not be able to sign transactions with the user's private keys.

I am not adopting that change as-is and will investigate the certificate/TLS architecture separately.