Improve the integration of WalkerLog classes in batched drivers

[ye-luo]

Proposed changes

Made the following changes.

1. WalkerLogCollector now captures 'state' during construction. No more copy and update. This modification also requires WalkerLogCollector created via WalkerLogManager which provides 'state'.
2. WalkerLogManager acquires a list of references of WalkerLogCollector from crowds at startRun and releases the reference at stopRun. Not using pointers directly any more
3. wlog_collector_ are uniuqe_ptr in Crowd. I would like to eventually moving away it way from state.logs_active a deeply hidden state.
4. wlog_manager is a function scoped object now.

What type(s) of changes does this code introduce?

• Refactoring (no functional changes, no api changes)

Does this introduce a breaking change?

• No

What systems has this change been tested on?

epyc-server

Checklist

• Yes. This PR is up to date with current the current state of 'develop'
• Yes. Code added or changed in the PR has been clang-formatted

[PDoakORNL]

Efficient improvement. The .release() into the legacy drivers is a nice trick. Think this just needs the noted. documentation addition.

[ye-luo]

@jtkrogel could you review?

[jtkrogel]

Just one addition requested below (logic check).

Overall, the change to a "capture" style for collectors makes the WalkerLogManager function API ambiguous due to semi-ownership of the collectors. Functions accepting collectors or not seems like an arbitrary choice, potentially raising questions as to whether the collectors passed in to a member function should ever differ from the captured set (they shouldn't). Despite this, I don't think its a blocking issue.

[ye-luo]

writebuffer requires the same set of collectors as startRun. Since the code has be changed to capture the set of references at the startRun, writebuffer doesn't need to take the set again as argument any longer.
startRun stopRun and writeBuffer are the only three public APIs. With PR, only startRun takes a set of references to collectors. So less error-prone.
Regarding private functions, I tried to leave the set of references to collectors as an argument if it helps making the function const. Such functions are more like free functions preferably not pulling member variables directly for exmaple checkCollectors. Regarding the other private member functions openFile openHDFFile, I feel there is some design but the implementation is incomplete and thus leave them as they are.

[ye-luo]

Test this please

[jtkrogel]

Good enough.