wifi: mt76: mt7925: track valid BSS txpower

[Komzpa]

Summary

Add one follow-up Wi-Fi patch for mt7925 TX power reporting before mac80211 has delivered BSS_CHANGED_TXPOWER.

Current patch 18 stores per-BSS TX power in dev->phy.txpower, but there is no validity state for that field. Before the first BSS_CHANGED_TXPOWER event, dev->phy.txpower is just the zero-initialized value. mt7925_mcu_set_rate_txpower() treats that zero as configured power, feeds it into mt76_get_power_bound(), and userspace reports txpower 0.00 dBm after reload/disconnected state.

This patch tracks whether BSS TX power has actually been received:

1. add txpower_set next to txpower;
2. set it when handling BSS_CHANGED_TXPOWER;
3. use dev->phy.txpower only when txpower_set is true, so a legitimate configured 0 dBm remains valid;
4. before BSS TX power is known, fall back to hw->conf.power_level, then current channel regulatory max power;
5. continue through mt76_get_power_bound() and mt76_get_rate_power_limits().

Verification

• patch --dry-run --fuzz=0 passes against the generated mt76 tree after patches 1-19.
• make sources passes on this branch.
• Generated source contains the expected txpower_set guard and channel max fallback.
• Live nucat evidence from the fuller test branch: after DKMS install and module reload without reboot, iw dev wlp8s0 info changed from txpower 0.00 dBm to txpower 23.00 dBm.

Scope and routing

This PR is intentionally minimized to the DKMS/backport txpower reporting fix only. It does not claim to fix Bluetooth/Wi-Fi coexistence and does not include the separate Linux 7.1 ieee80211_mgmt.u.action build-compat patch.

This is not a Linux wireless upstream submission as-is: current wireless-next does not have this repo's patch-18 dev->phy.txpower path, so a real upstream submission would need a separate mt76 patch/series against wireless or wireless-next and should be sent via git send-email to linux-wireless rather than as this GitHub patch-file PR.