[thaliamontreux]

Description

Fixes #3857. cape_cin converts the parcel and environment profiles to virtual temperature, then passes them to lfc/el, which computed the LCL from that virtual temperature as though it were air temperature. Since Tv >= T, the LCL came out at too low a pressure (too high an altitude), which could in turn push the LFC and EL too high.

Adds an optional keyword-only parcel_temperature_start parameter to both lfc and el (paired with the existing dewpoint_start), used only for the LCL determination; defaults preserve existing behavior (parcel_temperature_profile[0] for lfc, temperature[0] for el). cape_cin now passes the parcel's true starting temperature, keeping the buoyancy comparisons in virtual-temperature space while constraining the LCL correctly.

On the #1190 regression sounding this corrects the LCL/LFC from 866.9 to 900.1 hPa and CAPE from 2161.9 to 2206.7 J/kg. The one updated expected value (test_cape_cin_value_error) was verified against the physical mechanism.

Checklist

• Closes Virtual temperature being incorrectly used in lfc and el functions #3857
• Tests added (test_lfc_parcel_temperature_start) and the cape_cin ValueError: zero-size array to reduction operation minimum which has no identity #1190 regression value updated
• Fully documented (docstrings for the new parameter)
• Backwards compatible: new parameter is keyword-only and optional

[thaliamontreux]

Closing as a duplicate of #4067, which fixes #3857 via a different approach (lcl_pressure/lcl_temperature kwargs passed from cape_cin) and predates this PR. For what it's worth, I verified the underlying mechanism independently: on the #1190 regression sounding the virtual-temperature LCL comes out at 866.9 hPa vs the true 900.1 hPa, so the fix direction is confirmed either way.