Running out of inodes during "pmbootstrap install"

For some device + UI combinations, copying data into the image at the end of pmbootstrap install fails with errors like (job):

[09:22:24] (native) % cd /mnt/rootfs_pine64-pinebookpro; cp -a usr lib run sys srv tmp mnt var media root boot dev sbin proc opt bin etc /mnt/install/
cp: can't create '/mnt/install/etc/skel/.profile': No space left on device
cp: can't create directory '/mnt/install/etc/skel/.config': No space left on device
cp: can't create directory '/mnt/install/etc/signon-ui': No space left on device
cp: can't create '/mnt/install/etc/fstab': No space left on device
cp: can't create '/mnt/install/etc/doas.conf': No space left on device

I have mostly seen this with plasma mobile and plasma desktop, for example: master:pine64-pinebookpro:plasma-desktop. But this doesn't mean it is the fault of these UIs, these just happen to install a lot of small files.

df -i shows that we are running out of inodes:

build:~$ df -i -h
Filesystem     Inodes IUsed IFree IUse% Mounted on
devtmpfs         487K   185  487K    1% /dev
shm              491K     2  491K    1% /dev/shm
/dev/vda3        1.4M  120K  1.3M    9% /
tmpfs            800K    68  800K    1% /run
/dev/vda1         25K    25   25K    1% /boot
tmpfs            491K    22  491K    1% /home/build/.local/var/pmbootstrap/chroot_native/dev
tmpfs            491K     1  491K    1% /home/build/.local/var/pmbootstrap/chroot_native/dev/shm
/dev/installp2    98K   98K     0  100% /home/build/.local/var/pmbootstrap/chroot_native/mnt/install
/dev/installp1    58K   124   57K    1% /home/build/.local/var/pmbootstrap/chroot_native/mnt/install/boot

Looking at pmb/install/format.py, we actually hardcode the ~100K inodes currently.

This has been previously discussed here years ago:

• https://gitlab.com/postmarketOS/pmbootstrap/-/issues/1717
• https://gitlab.com/postmarketOS/pmbootstrap/-/issues/1845

The key takeaway was that we do need to specify an amount of inodes as the default is not enough. But the default we used earlier is clearly too small, I'll prepare a patch to increase it.

Created this issue so we don't need to reference the old ones in the source anymore.