backends(cuda/ref,hip/ref): fix use of incorrect output field index

minor - update copyright to 2026

atpoints - fix insidious memory bug

style: Change clang-format to penalize newline after open ()

gpu & at-points - fix bug in output vectors for diagonal assembly

gpu - add state checking for at-points element restriction

at-points - ensure reference point coordinates are always up to date

api - naming consistency for composite CeedOperator

Pre-request work vectors in operator setup

Reset work vectors in operator setup to save memory

op - gpu minor at points diagonal improvements

Fix edge case in offset padding for empty cells

minor - simpler upper bound

vec - make CopyStrided reflect SetStrided

gpu - fewer threads for SetValueStrided

vec - drop unused length arg in setValueStridedK

vec - add stop to CeedVectorSetValueStrided

minor - upate copyright to 2025

gpu - allow running shared kernels on stream

gpu - prefer cu/hipBlas over handrolls

gpu - minor consistency

vec - use memset on GPU when SetValue for 0

Vector API compliance for CUDA backends

cuda - nontensor gen operators

gpu - split AtPoints basis between Transpose/no