handling non-numeric TRES

[paciorek]

When a TRES value is not numeric (e.g., mem=512M), sq errors:

(Pdb) qos_df['GrpTRES']
102    cpu=1120,mem=5151200M
Name: GrpTRES, dtype: object
(Pdb) qos_df['GrpTRES'].values[0]
'cpu=1120,mem=5151200M'
(Pdb) parse_tres(qos_df['GrpTRES'].values[0])
*** ValueError: invalid literal for int() with base 10: '5151200M'

I think a simple fix is to use str() instead of int() in parse_tres:

tres[key] = str(value) #  int(value)

I don't think the values produced by parse_tres really need to be numeric, but it would be good to think more about it.