Parametric DMD for CSV database

[siuwuncheung]

LIST_DIR/TRAIN_LIST.csv and LIST_DIR/TEST_LIST.csv:

• Both files are now assumed to have m rows and n+1 columns
• m is the number of {training, testing} parameter points
• n is the dimension of the parametric space
• first column is the name of the dataset DATASET, where the actual dataset is located in DATA_DIR/DATASET
• the second column through the last column denote n (could be zero) parameter values

OUTPUT_DIR/indicator_val.csv:

• Optional. Present only for time windowing cases.
• It is now placed in the output directory, since it depends on the snapshot used and the model parameter nwinsamp.
• It can be user-defined or auto-generated.

local_tw_csv.cpp:

• Renamed from generic_csv.cpp
• It is now limited to local (time-windowing) DMD.
• Only allows one training parameter, i.e. only 1 row in LIST_DIR/TRAIN_LIST.csv.
• Allow multiple testing parameters predicted in one single run, i.e. multiple rows in LIST_DIR/TEST_LIST.csv.
• Ignore the parameters values, i.e. the second column through the last column of LIST_DIR/TRAIN_LIST.csv and LIST_DIR/TEST_LIST.csv.
• Prediction is made after training DMD models with -train (default) or use trained-and-saved models with -no-train.
• As in the master branch, Indicator range partition and classification of snapshots is done simultaneously during the training of DMD.

parametric_tw_csv.cpp:

• Newly added for parametric (time-windowing) DMD.
• Offline and online phases are split.
• (Offline & online) Only allows multiple training parameters, i.e. more than 1 rows in LIST_DIR/TRAIN_LIST.csv.
• (Online) Allow multiple testing parameters predicted in one single run, i.e. multiple rows in LIST_DIR/TEST_LIST.csv.
• (Offline) Time windowing procedure: Indicator range partition is decided -> snapshots are classified into windows for each parameter -> training of DMD for each window and parameter.
• (Online) Interpolating DMD models using the parameters values, i.e. the second column through the last column of LIST_DIR/TRAIN_LIST.csv and LIST_DIR/TEST_LIST.csv.
• (Offline & online) Prediction is made after training DMD models with -predict (default), on the training/testing datasets in the offline/online phase.

Other changes:

• Add an example script on heat conduction problem heat_conduction_csv.sh in examples/dmd
• Add results to the top of the DMD-CSV scripts local_tw_csv.cpp and parametric_tw_csv.cpp in examples/dmd

[siuwuncheung]

Thanks @dylan-copeland for reviewing. I will add more changes to this PR, so I remove RFR and add WIP back. I will review the code as suggested, before moving to RFR again.

[siuwuncheung]

After some major changes, this PR is now ready for review.

[siuwuncheung]

@dylan-copeland has finished the review and informed me that the PR is ready to merge.