fix(ci,rtx-tensor): resolve clippy --all-features intel-mkl conflict; gate MKL to x86_64-linux

clippy --all-features enabled both rtx-tensor/mkl (intel-mkl-src mkl-static-lp64-seq)
and rtx-csm/candle mkl (mkl-static-lp64-iomp) -> two conflicting intel-mkl-src link
configs -> E0428 'MKL_CONFIG defined multiple times'.

- clippy: drop --all-features (lint default features; --all-features is unsound for a
  multi-platform, mutually-exclusive-backend workspace).
- rtx-tensor: gate intel-mkl-src to cfg(all(target_os=linux, target_arch=x86_64)) so
  mkl is never pulled on macOS/arm.

Co-Authored-By: Claude Opus 4.8 <[email protected]>

.gitea/workflows/ci.yml

@@ -32,7 +32,7 @@ jobs:
     - uses: Swatinem/rust-cache@v2
     - name: Run Clippy
       run: |
-        cargo clippy --workspace --all-features -- \
+        cargo clippy --workspace -- \
           -W clippy::all \
           -W clippy::pedantic \
           -A clippy::module_name_repetitions \

crates/core/rtx-tensor/Cargo.toml

@@ -49,9 +49,9 @@ blas-src = { version = "0.10", default-features = false, optional = true }
 # NOTE: Requires gfortran to be installed. Currently disabled by default due to ARM64 issues.
 openblas-src = { version = "0.10", features = ["cblas", "system"], optional = true }
 
-# Intel MKL backend (optional) - best CPU performance for large matrices
+# Intel MKL backend (optional, x86_64-Linux only - see target section below).
-# Enable with: cargo build --features mkl
+# MKL has no Apple Silicon build and collides with candle's mkl variant, so it is
-intel-mkl-src = { version = "0.8", features = ["mkl-static-lp64-seq"], optional = true }
+# gated to x86_64 Linux. Enable with: cargo build --features mkl (on x86_64 Linux).
 
 # Real CUDA GPU acceleration with cudarc 0.18.1 for production RTX 5090 support (matching workspace version)
 cudarc = { workspace = true, optional = true }
@@ -97,6 +97,11 @@ objc2-metal-performance-shaders = { version = "0.3", optional = true, features =
     "MPSKernel"
 ]}
 
+# Intel MKL is x86_64-Linux only (no Apple Silicon build); gate it so the `mkl`
+# feature can never pull it on macOS/arm and collide with candle's mkl variant.
+[target.'cfg(all(target_os = "linux", target_arch = "x86_64"))'.dependencies]
+intel-mkl-src = { version = "0.8", features = ["mkl-static-lp64-seq"], optional = true }
+
 [features]
 default = ["cpu"]  # No BLAS by default - manual impl works on all platforms
 cuda = ["cudarc"]