Improve the object code in generate_call_stub

Fix generate_fast_get_int_field0 on jniFastGetField_riscv32.cpp

Fix pass double args in jniTypes_riscv32.hpp

Make load_at/store_at atomically in barrierSetAssembler

Co-authored-by: zifeihan <caogui@iscas.ac.cn>

[rv32g-dev] Remove assertions not satisfied in current jdk version

Fix x11 registers not saving correctly when using synchronized

Fix pass long args in jniTypes_riscv32.hpp

Fix configure integration

With reference to https://bugs.openjdk.java.net/browse/JDK-8283020, there is the same issue on the riscv32 platform.
This patch will remove some modification about autoconf-* files due to legal reasons.

Fix the slli instruction, and the offset when the sp register is moved.

Co-authored-by: zhangxiang-plct <zhangxiang@nj.iscas.ac.cn>

Fix local variable offset error in generate_normal_entry function

Fix pr170 on PlatformLoad and PlatformStore

Fix return long type in generate_call_stub function

Fix dconst in templateTable_riscv32.cpp

* The current instruction only works with single-precision 0, so modified to load double-precision 0 from memory.

Fix convert of long/double/float

Expand BUFFER_SIZE in jniFastGetField_riscv32.cpp

Fix BUFFER_SIZE in jniFastGetField_riscv32.cpp to fix exceptions in codeBuffer.cpp when the release version is running.

* Use SharedRuntime to implement convert of long/double/float.

Rv32g dev javac

Fix lshl/lshr/lushr in templateTable

* Only the low 6 bits of rs2 are considered for the shift amount.

Use int32 to instead some int64

Co-authored-by: zhangxiang-plct <zhangxiang@nj.iscas.ac.cn>

Co-authored-by: zhangxiang-plct <zhangxiang@nj.iscas.ac.cn>

* Refer to the 32-bit ldiv implementation of other architectures to fix ldiv in rv32.