fix: avoid useless io in large-payload queries

[zghong]

Description

When executing large-payload insert queries by POST, the current retry and throughput-limit mechanisms read the POST body, which can be resource-consuming and may even cause CPU/Memory bottlenecks. This MR avoids unnecessary I/O operations in these two mechanisms:

• For the retry mechanism: If the retry parameter is not set, the step of copying the body can be skipped.
• For the throughput-limit mechanism: Instead of calculating the length of query, the Content-Length is read directly from POST header.

Closes #499, #428

Pull request type

Please check the type of change your PR introduces:

• Bugfix
• Feature
• Code style update (formatting, renaming)
• Refactoring (no functional changes, no api changes)
• Build related changes
• Documentation content changes
• Other (please describe):

Checklist

• Linter passes correctly
• Add tests which fail without the change (if possible)
• All tests passing
• Extended the README / documentation, if necessary

Does this introduce a breaking change?

• Yes
• No

Further comments

For the throughput-limit mechanism: The Content-Length header indicates the size of the request body, so the throughput of GETs will be ignored. This is logical, as GETs typically have minimal impact on throughput compared to POSTs.

[zghong]

Before:

After:

[mga-chka]

nice improvement