* Added multiple `partprobe` calls and added a `.partprobe()` function on partitions, filesystem and blockdevice.
* Adding retry attempts to all UUID related operations tied to the boot process
* Tweaked logging for mounting and disk related operations
* Removed potential SysCall exception disruptor causing exceptions to go by unnoticed
* Increased the start position from 1MiB to 5MiB of /boot partition
* Optimized the GRUB installation & config code
* Improved Partition().uuid to never return None. Instead it will raise an exception if it can't get a PARTUUID within X retries with Y delay per attempt.
* Increased sleep timer for partition uuid retrieval, because even with a 3 second sleep it wasn't long enough even on fast devices.
* Make GRUB install to /dev/sda instead of /dev/sda1.
* Added 10 retries for retreiving PARTUUID with a one second sleep. Instead of increasing the sleep simply add more retries until we find a good balance on slower disks.
* Adding partprobe at strategic places.
* Swapped `for partition in blockdevice` to `for uuid, partition in blockdevice.partitions.items()` instead as `__iter__` for debugging purposes.
* `get_mount_info()` now causes a exception rather than returning nothing if there is nothing to be shown. This to avoid issues where in places this is crucial information and it went by unnoticeable. Using exception handlers where it doesn't matter if there's any information or not.
Created a new JSON serializer called `UNSAFE_JSON` that will serialize everything, including sensitive information. And `JSON` which is the default up to this point now safely ignores any sensitive information in dictionaries marked with `!`, for instance `{"!password" : "mypassword"}` will be omitted from any output.
It seems the system does not syncronus update its internal information after a partitioning.
Two places are affected. Directly on filesystem.add_partition (the uuid of the new partition isn't available after the parted command)
and blockdevice.get_partition, where the list of partitions for the iterator might not be available in the query.
The patch places both sections under controlled loops, giving the system the chance to update the information.
Should be more controlled via application parameters