Implement generic abstract command support cache

[en-sc]

The idea is similar to #928.

The issue

Currently, there are many target properties derived from abstract command being supported or not:

riscv-openocd/src/target/riscv/riscv-013.c

Lines 178 to 183 in f32f178

	bool abstract_read_csr_supported;
	bool abstract_write_csr_supported;
	bool abstract_read_fpr_supported;
	bool abstract_write_fpr_supported;
	yes_no_maybe_t has_aampostincrement;

There are issues with the current approach, e.g.:

• Extensibility: Take abstract_..._supported for example. Current implementation assumes the support is the same across all the registers in a group.
• Correctness: We should make sure abstractcs.cmderr == not_supported, not abstractcs.cmderr != none before asserting that the support is missing. This is not currently done in some places, e.g.
  

  riscv-openocd/src/target/riscv/riscv-013.c

  Lines 3791 to 3807 in f32f178

  	/* TODO: we need to modify error handling here. */
  	/* NOTE: in case of timeout cmderr is set to CMDERR_NONE */
  	if (info->has_aampostincrement == YNM_MAYBE) {
  	if (result == ERROR_OK) {
  	/* Safety: double-check that the address was really auto-incremented */
  	riscv_reg_t new_address;
  	result = read_abstract_arg(target, &new_address, 1, riscv_xlen(target));
  	if (result != ERROR_OK)
  	return MEM_ACCESS_FAILED_DM_ACCESS_FAILED;
  	if (new_address == args.address + args.size) {
  	LOG_TARGET_DEBUG(target, "aampostincrement is supported on this target.");
  	info->has_aampostincrement = YNM_YES;
  	} else {
  	LOG_TARGET_WARNING(target, "Buggy aampostincrement! Address not incremented correctly.");
  	info->has_aampostincrement = YNM_NO;
  	}

Proposed solution

• Store all non-supported commands in a cache (a sorted array should suffice).
• Query if the command is known to be not supported before executing a command (bsearch() in the array).
• Add the command after execution into the array iff abstractcs.cmderr is not_supported.

[JanMatCodasip]

In my opinion, this is a very good proposal. I agree with both the general idea and the proposed approach for implementation.