dysk stalls ~10s on Arch Linux with Btrfs since kernel 6.16.4

On Arch Linux with Btrfs, dysk takes about 10 seconds before printing any output.
This affects all commands that actually enumerate mounts, e.g.:

```
time dysk -c fs+type+size+used
time dysk -a

```

…but not dysk --help, which is instant.
The delay started suddenly after upgrading from kernel 6.16.3 → 6.16.4.
It persists in 6.16.6. It happens with both dysk 3.0.0 and 3.1.0, so it seems kernel-related rather than a dysk version change.

### Environment

- Distro: Arch Linux
- dysk: 3.0.0-1 and 3.1.0-1 (both tested)
- Kernel: 6.16.6-arch1-1 (regression appears between 6.16.3 and 6.16.4)
- Filesystem: Btrfs root with quotas enabled, multiple subvolumes (@, @home, @log, @pkg)
- systemd: 257.9-1

### Steps to reproduce
1. On Arch with Btrfs and kernel ≥ 6.16.4, run:
`time dysk -c fs+type+size+used
`
2. Observe ~10 second stall before output.

### Expected behavior

Output should be displayed instantly, as it was before kernel 6.16.4.

### Actual behavior

Consistent ~10s delay, even with filters or --remote-stats=no.

**Notes**
```

- dysk --help is instant (so startup overhead is not the issue).
- df -h and lsblk remain instant.
- Disabling quotas (sudo btrfs quota disable /) may help, but not confirmed.
- Possibly due to how dysk probes Btrfs mounts via statfs/statvfs.
- This regression only appeared after the kernel update, but affects dysk specifically.

**Repro commands / timings**
$ time dysk -c fs+type+size+used
┌──────────────┬─────┬────┬────┐
│ filesystem │type │size│used│
├──────────────┼─────┼────┼────┤
│/dev/nvme0n1p2│btrfs│511G│ 49G│
│/dev/nvme0n1p1│vfat │1.1G│280M│
└──────────────┴─────┴────┴────┘

real 0m10.188s
user 0m0.000s
sys 0m0.010s

$ time df -h /
Filesystem Size Used Avail Use% Mounted on
/dev/nvme0n1p2 476G 45G 431G 10% /

real 0m0.007s
user 0m0.003s
sys 0m0.004s
```

Reproduce the delay with `dysk -c fs+type+size+used` and `dysk -a` on Arch Linux with Btrfs, comparing kernels 6.16.3 and 6.16.4 or later. Start by tracing the mount-enumeration path and the mentioned statfs/statvfs probing, then compare it with `df -h` and `--remote-stats=no`. Done means identifying the responsible operation and documenting or resolving the regression.

Written by the indexing model from the issue text.