fix(python): use nanosecond nonces on every signed HTTP endpoint

[Dusk1e]

Summary

The Rust SDK has one nonce source, make_nonce() in crates/api-rust/src/api/parts/mod.rs, and every signed method uses it:

pub fn make_nonce() -> u64 {
    std::time::SystemTime::now()
        .duration_since(std::time::UNIX_EPOCH)
        .unwrap()
        .as_nanos() as u64
}

BulkHttpClient.place_orders and BulkWebSocketClient agree — both use time.time_ns(). But four other signed endpoints on the same client divide it:

endpoint	nonce expression	unit
place_orders	time.time_ns()	ns
update_leverage	int(time.time_ns() / 100000)	100 µs
manage_agent_wallet	int(time.time_ns() / 1000000)	ms
whitelist_faucet	int(time.time_ns() / 1000000)	ms
request_faucet	int(time.time_ns() / 1000000)	ms

Two consequences:

The nonce jumps backwards. Sampled at one frozen instant, the four divided endpoints emit a value 100000x–1000000x below what every other signed call from the same account emits. Any account that has placed an order is then sending leverage, agent-wallet and faucet transactions far below its own high-water mark.

Dividing quantises the nonce. int(ns / 1000000) is millisecond-granular by construction, so two calls inside the same millisecond produce the same nonce — which is the one thing a nonce has to avoid.

Changes

Use time.time_ns() in all four, matching place_orders, the WebSocket client and make_nonce(). Explicit nonce= arguments are untouched, and the two stale "in milliseconds" comments are corrected.

Verification

Loading bulk_http.py with the stubs the repo's own tests use, freezing the clock and reading the nonce each endpoint puts on the wire:

clock frozen at time.time_ns() == 1786500000123456789
(Rust SDK make_nonce() would return 1786500000123456789)

                      before                             after
endpoint              nonce on the wire   vs Rust        nonce on the wire   vs Rust
------------------------------------------------------------------------------------
place_orders          1786500000123456789        1x      1786500000123456789      1x
update_leverage       17865000001234        100000x      1786500000123456789      1x
manage_agent_wallet   1786500000123        1000000x      1786500000123456789      1x
whitelist_faucet      1786500000123        1000000x      1786500000123456789      1x
request_faucet        1786500000123        1000000x      1786500000123456789      1x

distinct nonces from two calls 500,000 ns apart, inside one millisecond:
                      before              after
  manage_agent_wallet 1 (SAME NONCE)      2 (unique)
  whitelist_faucet    1 (SAME NONCE)      2 (unique)
  request_faucet      1 (SAME NONCE)      2 (unique)

tests/test_history_http.py and tests/test_wire_serialization.py are unchanged by this (20 passed either way; the history failures on main are #7).