Fix L2 norms in C_util/Convergence codes

[ajnonaka]

Summary

The multifab L2 norm routines simply return sum(a dot a). Thus, when you sum up norms from each grid you do NOT need to take them to the ^NORM power. The cell count (i.e. volume) scaling is performed lower

The formula is now implemented correctly, for p!=0:

L^p = ( 1/Ncell sum (phi_f - phi_c)^p ) ^(1/p)

(note that using cell volume instead of 1/Ncell accomplishes the proper scaling as well, which is what the code does)

Additional background

Checklist

The proposed changes:

• [x] fix a bug or incorrect behavior in AMReX
• [ ] add new capabilities to AMReX
• [ ] changes answers in the test suite to more than roundoff level
• [ ] are likely to significantly affect the results of downstream AMReX users
• [ ] include documentation in the code and/or rst files, if appropriate

[zingale]

note that it also worked for p =1

[zingale]

this is a nice catch. L2 never behaved well for me