Add hdf_archive dataset append feature

Then we can get rid of direct calls
```
../src/QMCDrivers/WalkerLogBuffer.h: h5d_append(f.top(), "data", file_pointer, buffer.dim(), dims, buffer.data());
../src/QMCHamiltonians/ObservableHelper.h: /// "file pointer" for h5d_append
../src/QMCHamiltonians/ObservableHelper.cpp: h5d_append(file.top(), "value", current, rank, mydims.data(), first_v + lower_bound);
../src/Estimators/TraceManager.h: h5d_append(f.top(), "traces", file_pointer, buffer.dim(), dims, buffer.data());
```

Start by tracing hdf_archive's dataset-writing entry points and the direct h5d_append calls in WalkerLogBuffer.h, ObservableHelper.h, ObservableHelper.cpp, and TraceManager.h. Determine the shared append behavior required by these callers; done means they no longer call h5d_append directly and the archive API supports their dataset appends.

Written by the indexing model from the issue text.