Inline harmonic analysis

[c2xu]

1. Currently, the harmonic analysis does not have the restart capability to span multiple run segments.
2. Array indexing should be more consistent with that in the model, e.g., using hor_index_type to specify the dimensions.
3. The use of harmonic analysis should be specified by a logic flag, instead of relying on parameters inside the module.
4. Revisit the use of local variable instead of pointer for the control structure.

[c2xu]

5. The subroutine HA_init is called from the subroutine tidal_forcing_init, but it seems that in regional models tides can be forced from open boundaries only without enabling astronomical tidal forcing.

[c2xu]

6. Right now only one accumulator of FtF is implemented (at each baroclinic time step) and is shared between all FtSSH. This would become problematic if some FtSSH is accumulated at different time steps (e.g., the barotropic time step).
7. Also, the harmonic analysis doesn't really need such a high temporal resolution. Instead of each baroclinic time step, they can be called at, e.g., each hour. It might be more convenient to have a user-specified sampling frequency.