fix(ai-prompt-*): handle nil err in JSON body parse path

[shreemaan-abhishek]

Description

Follow-up to #13096 (which corrected the wording of the JSON-parse error message in the AI prompt plugins). That patch left two nil-concatenation hazards behind that turn the affected plugins into 500s instead of returning the intended 4xx client error:

1. core.request.get_body() can return nil with err == nil for empty request bodies. With:

   return nil, { message = "could not get body: " .. err }

   the concatenation crashes with attempt to concatenate local 'err' (a nil value).

2. core.json.decode() returning nil with err == nil is much rarer but still possible, and the same concatenation hazard applies.

Affected files:

• apisix/core/request.lua — get_json_request_body_table(). The get_body() branch already had an (err or "request body is empty") guard from a previous patch; this PR adds the matching guard on the json.decode() branch.
• apisix/plugins/ai-prompt-template.lua — get_request_body_table(). Both branches need the guard.
• apisix/plugins/ai-prompt-decorator.lua — get_request_body_table(). Both branches need the guard.

Reproduction (before this PR):

POST / HTTP/1.1
Content-Type: application/json
Content-Length: 0

with ai-prompt-template enabled returns 500 instead of the configured "could not get body: request body is empty" message.

Which issue(s) this PR fixes:

Follow-up to #13096; no separate issue.

Checklist

• I have explained the need for this PR and the problem it solves
• I have explained the changes or the new features added to this PR
• I have added tests corresponding to this change — the existing t/admin/routes_request_body.t exercises the nil-err empty-body path and now correctly produces the configured error message; happy to add a focused regression test in t/plugin/ai-prompt-template.t if reviewers want one.
• I have updated the documentation to reflect this change — no doc impact; behaviour is now the documented one.
• I have verified that this change is backward compatible