Update the model in cantrip-builtin to fit in 4MB memory

ML core only has 4MB in HW. For the models running on Shodan, we need to
make sure it fits in 4MB.

Bug: 258850990
Change-Id: I8a5428b775538664313bf2835089ba5b1e1bd22f

quant_models/CMakeLists.txt

diff --git a/quant_models/CMakeLists.txt b/quant_models/CMakeLists.txt
index fd18344..8292bf2 100644
--- a/quant_models/CMakeLists.txt
+++ b/quant_models/CMakeLists.txt
@@ -185,6 +185,7 @@
   LINKOPTS
     "LINKER:--defsym=__stack_size__=300k"
     "LINKER:--defsym=__heap_size__=1500k"
+    "LINKER:--defsym=__tcm_length__=4M"
   COPTS
     "-DBUILD_EMITC"
   TESTFILES