implement hist_empty_htapes for mosart

[mahf708]

Task: Add an rtmhist_empty_htapes namelist option to MOSART

Goal: Add a namelist flag to MOSART (the river-routing component, components/mosart/) that suppresses the
default history fields, exactly mirroring ELM's existing hist_empty_htapes option. When set to .true., no
history field is added to a tape unless explicitly requested via rtmhist_fincl[1-3]. This lets users
disable MOSART history output entirely (by setting rtmhist_empty_htapes = .true. with no fincl), the same
way hist_empty_htapes = .true. works for ELM.

Background / why: MOSART currently has no equivalent to ELM's hist_empty_htapes. Its history code
(RtmHistFile.F90) is a near-verbatim copy of ELM's histFileMod.F90, so this is essentially porting one line
plus the namelist plumbing. The reference implementation in ELM is at
components/elm/src/main/histFileMod.F90:707 (else if (.not. hist_empty_htapes) then) and its plumbing in
components/elm/src/main/controlMod.F90 (declaration, namelist read, mpi_bcast). Use ELM as the template for
naming and style.

Exact changes required (all in components/mosart/):

1. src/riverroute/RtmHistFile.F90 — declare the module variable near the other rtmhist_* public namelist
   vars (around lines 44–60, e.g. next to rtmhist_fincl1):
   logical, public :: rtmhist_empty_htapes = .false. ! namelist: flag indicates no default history fields
2. src/riverroute/RtmHistFile.F90:371 — in subroutine htapes_fieldlist, the masterlist loop currently
   reads:
   else
   ! find index of field in exclude list
   call list_index (fexcl(1,t), mastername, ff)
3. Change the bare else to:
   else if (.not. rtmhist_empty_htapes) then
4. (This matches ELM histFileMod.F90:707.)
5. src/riverroute/RtmMod.F90 — three small additions in the namelist-reading routine:
   - Add rtmhist_empty_htapes to the use RtmHistFile, only: ... import list (around lines 30–32).
   - Add rtmhist_empty_htapes to the namelist /mosart_inparm/ declaration (around lines 271–274).
   - Add an mpi_bcast for it alongside the other rtmhist_* broadcasts (around lines 403–414):
   call mpi_bcast (rtmhist_empty_htapes, 1, MPI_LOGICAL, 0, mpicom_rof, ier)
6. bld/namelist_files/namelist_definition_mosart.xml — add an entry near the other rtmhist_* entries (lines
   248–335) so the option validates:
   
   If TRUE, indicates do not output any default history fields. Only fields
   explicitly requested via rtmhist_fincl[1-3] will be written. Mirrors ELM's
   hist_empty_htapes.

5. build-namelist — no change required. rtmhist_fincl1/rtmhist_fexcl1 already flow through user_nl_mosart without an explicit add_default call in bld/build-namelist, so the new entry behaves the same; the Fortran default .false. covers the unset case. (Optional: add .false. to bld/namelist_files/namelist_defaults_mosart.xml for explicitness, but it is not necessary.)

Important correctness note to verify: With rtmhist_empty_htapes = .true. and no fincl, tape 1 ends up with
nflds == 0. In RtmHistFile.F90:418-424, ntapes is computed by scanning from max_tapes down for the first
tape with nflds > 0, so it resolves to 0. The "no empty tapes / no holes" abort loop at RtmHistFile.F90:429
is do t = 1, ntapes, which then no-ops (does not abort). All downstream history loops are also do t = 1,
ntapes, so they no-op. This is the same path ELM relies on when hist_empty_htapes = .true.. Confirm by
reading the history-write and restart-write paths in RtmHistFile.F90 that nothing assumes ntapes >= 1.

Constraints:

• Match existing code style/indentation in each file. Do not reformat surrounding code.
• Do not rename or alter the existing rtmhist_* options.
• Keep the default .false. so existing behavior is unchanged when the option is not set.

Validation:

• Confirm it compiles (build a case that includes MOSART, e.g. a WCYCL compset).
• Functional check: in a test case's user_nl_mosart, set rtmhist_empty_htapes = .true. and confirm MOSART
  writes no .mosart.h.nc history files (only restart files), while the run completes normally. Also confirm
  that setting it .true. together with an explicit rtmhist_fincl1 = 'RIVER_DISCHARGE_OVER_LAND_LIQ' produces
  a tape with only that one field.
• Run an ERS test on a MOSART-active compset to confirm restart behavior is unaffected (MOSART writes its
  own restart/history, so this exercises the ntapes == 0 path across a restart).

Deliverable: Report the diff of all files changed, and the result of the compile + functional checks.