crashdump reserves too much memory on small systems (and maybe not enough on large systems)

[ddstreet]

The current reservation is hardcoded at 256m, which can render smaller systems unusable due to reserving too much memory, e.g. Azure B1 instance is only 1G and B1ls is only 500m. On the other end, 256m may not be enough memory for systems with large amounts of memory, e.g. some Azure D* or E* instance types go up to 512G (or more).