fix(sipnet): use max timesteps per day to detect out_day, fixing partial first-day edge case

[anshul23102]

Summary

Fixes #3624.

model2netcdf.SIPNET determines how many output timesteps SIPNET writes per day (out_day) by counting rows for a single day. The old code always used the first day of the simulation year:

out_day <- sum(
  sipnet_output$year == simulation_years[1] &
    sipnet_output$day == unique(sipnet_output$day)[1],
  na.rm = TRUE
) # switched to day 2 in case first day is partial

The comment acknowledged the problem (partial first day) but the code still read from day 1 ([1]), not day 2. If SIPNET started mid-day, day 1 would have fewer rows than a complete day, making out_day too small and applying an incorrect timestep to all flux unit conversions for the entire year.

The fix counts rows per day across all days in the first simulation year and takes the maximum, which is the correct count for a complete day regardless of whether the first or last day is partial:

steps_per_day <- table(
  sipnet_output$day[sipnet_output$year == simulation_years[1]]
)
out_day <- max(steps_per_day)

Changes

• models/sipnet/R/model2netcdf.SIPNET.R: replace single-day row count with max(table(...)) over all days
• models/sipnet/tests/testthat/test-model2netcdf.SIPNET.R: add regression test with a partial first day (1 timestep on day 1, 2 on days 2 and 3) that would have produced wrong GPP values under the old code
• models/sipnet/NEWS.md: changelog entry

Test plan

• New test "out_day is correct when the first day of output is partial" passes
• Existing SIPNET tests still pass (normal complete-day runs are unaffected)