Implementing cpu die_completion from cpu_specs

[da-ekchajzer]

Implement a new way of completing die_size (Linked to #199)

• We directly use die_size and no longer die_size_per_core
• We no longer use cpu_manufacturer but only cpu_specs
• The completion strategy work as follows :

if die_size_per_core and core_units are given :

die_size = core_units*die_size

Otherwise, if family is given, die_size can be retrieved from a fuzzy matching on the cpu_specs repository.

If several cpu matches the given family, we use the core_units attributes :

• If core_units matches one to many cpu, the average value is given and min and max value are used as min and max fields.
• If core_units does not match any cpu, we infer the die_size with a rule of three or a linear regression (when multiple cpus are available).
• If core_units is not provided, the average value is given and min and max value are used as min and max fields.