Enable compilation for specific target

By exposing the target information through `CompilerConfig`,
compiler(only LLVM at the moment) could create a machine with
different CPU feature flags other than current host, which makes it
capable to "cross compile" to some degree.

Update #959

lib/runtime-core/src/backend.rs

@@ -129,6 +129,11 @@ pub struct CompilerConfig {
     pub enforce_stack_check: bool,
     pub track_state: bool,
     pub features: Features,
+
+    // target info used by LLVM
+    pub triple: Option<String>,
+    pub cpu_name: Option<String>,
+    pub cpu_features: Option<String>,
 }
 
 pub trait Compiler {

lib/runtime-core/src/codegen.rs

@@ -74,6 +74,13 @@ pub trait ModuleCodeGenerator<FCG: FunctionCodeGenerator<E>, RM: RunnableModule,
     /// Creates a new module code generator.
     fn new() -> Self;
 
+    /// Creates a new module code generator for specified target.
+    fn new_with_target(
+        triple: Option<String>,
+        cpu_name: Option<String>,
+        cpu_features: Option<String>,
+    ) -> Self;
+
     /// Returns the backend id associated with this MCG.
     fn backend_id() -> Backend;
 
@@ -206,7 +213,14 @@ impl<
             validate_with_features(wasm, &compiler_config.features)?;
         }
 
-        let mut mcg = MCG::new();
+        let mut mcg = match MCG::backend_id() {
+            Backend::LLVM => MCG::new_with_target(
+                compiler_config.triple.clone(),
+                compiler_config.cpu_name.clone(),
+                compiler_config.cpu_features.clone(),
+            ),
+            _ => MCG::new(),
+        };
         let mut chain = (self.middleware_chain_generator)();
         let info = crate::parse::read_module(
             wasm,