Use FFT Poisson solver on fully refined levels

[ChunYen-Chen]

Resolve Issue

Resolved #446.

Updates

Major

• Use FFT Poisson solver on fully refined levels
• Refactor Init_FFTW() and End_FFTW()

Minor

• Avoid hard-coded level 0 in src/Model_ELBDM/CPU_ELBDM/CPU_ELBDMSolver_FFT.cpp and src/Output/Output_BasePowerSpectrum.cpp

Tests done

Gravity

1. Change NX0_TOT_* = 32 and MAX_LEVEL = 7 in Input__Parameter
2. Add a new criteria in Input__Flag_Rho to refine everything on level 0

   # Level                         Density
         0                       1.0000e-16
         1                       1.0000e-2
         2                       4.0000e-2
         3                       1.6000e-1
         4                       6.4000e-1
         5                       2.5600e+0
         6                       1.0240e+1
         7                       4.0960e+1
         8                       1.6384e+2
         9                       6.5536e+2
   

3. Run the test problem
4. The solved potential is the same as the original run.

JeansInstability

1. Change NX0_TOT_* = 32, MAX_LEVEL = 1, and OPT__FLAG_RHO = 1 in Input__Parameter
2. 2. Add a file Input__Flag_Rho to refine everything on level 0

   # Level                         Density
         0                       1.0000e-16
   

3. Run the test problem
4. The results are the same as the original run.