fix: add compile-time get overloads for RVV to resolve ambiguity

RVV only had runtime get(batch, size_t, requires_arch<rvv>) which
became ambiguous with the new compile-time get(batch, index<I>,
requires_arch<common>) because index<I> (std::integral_constant)
implicitly converts to size_t. Add index<I> overloads that delegate
to the runtime versions, matching the pattern used by SSE/AVX/NEON.

Author: DiamonDinoia

include/xsimd/arch/xsimd_rvv.hpp

@@ -1370,6 +1370,19 @@ namespace xsimd
             return std::complex<T> { detail::rvvmv_lane0(tmpr), detail::rvvmv_lane0(tmpi) };
         }
 
+        // get (compile-time index)
+        template <class A, size_t I, class T, detail::rvv_enable_all_t<T> = 0>
+        XSIMD_INLINE T get(batch<T, A> const& arg, index<I>, requires_arch<rvv>) noexcept
+        {
+            return get(arg, I, rvv {});
+        }
+
+        template <class A, size_t I, class T, detail::rvv_enable_all_t<T> = 0>
+        XSIMD_INLINE std::complex<T> get(batch<std::complex<T>, A> const& arg, index<I>, requires_arch<rvv>) noexcept
+        {
+            return get(arg, I, rvv {});
+        }
+
         // all
         template <class A, class T, detail::rvv_enable_all_t<T> = 0>
         XSIMD_INLINE bool all(batch_bool<T, A> const& arg, requires_arch<rvv>) noexcept