Some code moved around to group together more logically.

Wasted most of a day finding them. :(

NOT COMPLETE! CURRENTLY EXITS AFTER FIRST W3SPR4 call
Made change to allow first W3SPR4 call (when IT=0) to run using all
arrays:
  - Updates some local variables with CHUNKSIZE dimension.
  - Some input parameters to W3SRCE now have extra NSEA[L[M]] dimension
  - Some chunk element loops (CSEA) added to blocks of code
  - W3SRCE only called once, rather than in seapoint loop

Chunk dimensions added to VSIN, VDIN and associated variables
Explicit loop added around calls to W3SINx routines in integration loop

Changes are currently B4B with develop (ST4 compile)

  - variables were being zeroed in chunk loop
  - claculations where being skipped as SRC_MASK was all true
Note: Had to recalculate the source mask after integration loop; I need
to find a better way of doing this.

Commented out some unusued vars

…_refactor

…after non-b4b results with non-homeogenous depth).

Also updated PHICE.

  - Required deprecation of REFLEC and REFLED locals in w3srce
  - Calculation of `REFLEC(4) * BERG` moved into w3srce
  - Needs #ifdefs in call to w3srce as REFLC and REFLD are not allocated
	if W3_REF1 not set
  - Had to make IX and IY chunk arrays in w3srce

  - Local scalars no longer used in w3wave
  - Requires #ifdef in W3SRCE call signature.

They are only used for PDLIB implicit source calls; avoids unneccesary
array allocation for direct souce calls.

energy in spectrum in w3str1

…loop).

Also removed some old code and unused variables.

…3SRCE

  - EMEAN had wrong index variable
  - D_INP needs ISEA in Miche Limiter calculation

pre implicit call. They are causing problems as they are not allocated.

… of NSEAL

during regression testing.

Might want to change this at some point to use chunked array (ST2 only)
Also added new code to write local CHUNKED arrays back to full seapoint
grid at end of chunk loop - some variables like USTAR are returned back
to w3wave.

calculation

Also calculate DTDYN correctly for implicit solver

Zeroing them element-wise later on.
Was causing B4B reproducibility issues in regtests.

Added some comments

…t pre calculation

elements in chunk have been completed before exiting.
Moved initialisaton of WHITECAP outside of chunk loop (was zeroing
previous calculations)

reproducibility.

compatibility with develop

…_refactor

to set tile/chunk size in new refactored W3SRCEMD:W3SRCE routine.

…_refactor

W3SRCE call.

…_refactor

when slicing.

…_refactor

…_refactor

…_refactor

…_refactor