Bump mlir-air to e279756 (fix i8 matmul compile time regression)

Update mlir-air dependency to include Xilinx/mlir-air#1531, which fixes
column-aware L2 memref-to-memtile assignment in the air-to-aie pass.

The blind round-robin L2 buffer placement caused cross-column routing
flows for large i8 matmul kernels (8x4 herd), making
aie-create-pathfinder-flows spend ~2m20s instead of <0.5s per
compilation. This caused the matmul_i8_m128_n64_k64 example to time
out in CI (36 invocations × ~2.5min each > 20min timeout).

With this fix, the full test suite completes in ~2m42s (cold cache).

Fixes #50

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: erwei-xilinx

utils/mlir-air-hash.txt

@@ -1,3 +1,3 @@
-Commit: e8c63ed
-Timestamp: 2026040805
+Commit: e279756
+Timestamp: 2026041318
 Version: 0.0.1