Claude review: scsi: ufs: exynos: stop poking into struct phy guts

[Claude Code Review Bot <claude-review@example.com>]

Patch Review

Replaces `generic_phy->power_count` checks with a local `ufs->phy_powered_on` bool. The commit message provides excellent analysis of why the original pattern was wrong (other consumers could affect power_count, and the read was unlocked).

The implementation correctly tracks state transitions in all paths:
- `exynos_ufs_phy_init()`: sets `false` before power-off, sets `true` after power-on
- `exynos_ufs_exit()`: early return if not powered on
- `exynos_ufs_suspend()`: sets `false` after power-off
- `exynos_ufs_resume()`: adds error handling for `phy_power_on()`, sets `true` on success

One minor note: in `exynos_ufs_resume()`, the function continues executing (`exynos_ufs_config_smu`, `exynos_ufs_fmp_resume`) even if `phy_power_on()` fails. The error is logged but not propagated to the return value. This seems intentional -- the original code didn't check the return either -- but it's worth noting that the resume function will return 0 even on PHY power failure. This preserves existing behavior, so it's acceptable.

---
Generated by Claude Code Patch Reviewer