fix!: restore clearnet price fetch when Mixnet Mode is switched off

[dorianvp]

This PR restores the price fetch when nym is switched off, which was not working.

[zancas]

The review finished with ten verified findings. I'll report them through the findings UI, then summarize.

● Code review(high · 10 findings)
zingolib/src/lightclient/mixnet.rs
● 1159 [privacy] clearnet_price_race collects all nine RACED_SOURCES into an unbounded FuturesUnordered, so the IP-disclosing route fans out nine-wide while the mixnet route is capped at SURVEY_WAVE_WIDTH (4).
● 1223 [test-coverage] switched_off_mode_consents_to_a_clearnet_fetch is a plain #[tokio::test] firing nine live HTTPS requests, and its empty Err(PriceError) arm accepts exactly the failure a routing regression produces.
● 1107 [correctness] The ConduitDial guard is taken before the acquirer branch that never uses it (then shadowed by a String named dial), so a spawned-session price race pins the standing conduit's in_flight count for a race travelling a freshly drawn exit.
● 1154 [duplication] clearnet_price_race restates PriceRun's declared policy and re-implements run_wave's drive-and-settle loop inline, leaving zingo-price's first_quote doc ('the racing lives in one place') false.
● 1093 [doc-comments] Six diff-authored item doc-comments are two sentences, violating the governing hard rule that every /// is exactly one sentence with no exceptions.
● 1104 [imports] All five diff-authored PriceFetchRoute use sites are inline full paths though both files already import siblings from those exact modules, violating the partial-imports rule.
● 1165 [efficiency] The clearnet legs reuse REQUEST_TIMEOUT (20s) and CONNECT_TIMEOUT (10s), bounds sized against the mobile watchdog for a tunnel-traversing fetch, so a switched-off fetch on a dead or captive-portal network holds update_current_price 10-20 seconds.
zingo-cli/src/commands.rs
● 2151 [stale-docs] Six surviving statements of the reversed mixnet-only price rule contradict the PR's behavior, led by the CurrentPrice long_about in the very file the PR edits.
docs/agents/net-diag-design.md
● 217 [breaking-change] The design doc says via_socks5 on the price-fetch success payload is consumed by zingo-mobile and must not be disturbed, and the PR removes that field without updating the doc or naming the mobile migration.
zingo-netutils/src/socks5_fetch.rs
● 157 [api-surface] Dropping #[cfg(feature = "testutils")] makes fetch_text_untunneled pub to every socks5-fetch consumer, so fail-closed transitional states are protected only by one match in update_current_price rather than the compiler.