Update wiki for split-config restructure (#140)

The config restructure in #140 deleted the monolithic YAMLs (base*.yaml, linear-dinov3-*.yaml, combined_*.yaml, concept_mermaid.yaml, benchmark.yaml) and the run1_*.csv source lists, replacing them with four split configs (data_config.yaml, model_config*.yaml, training_config*.yaml, logger_config.yaml). Several wiki pages still document the old layout and need updating.

wiki/Reproducibility.md references a run1_segformer_baseline.yaml run config with train_csv: configs/run1_train_sources.csv / val_csv: configs/run1_val_sources.csv — these files no longer exist; source whitelists are now inline whitelist_sources lists inside data_config.yaml. wiki/Writing-Tickets.md similarly points at configs/run1_segformer_baseline.yaml in its checklist. Both pages should be rewritten to show the four-file split-config model and the new scripts/train.py CLI (--config-data/--config-model/--config-training/--config-logger, each with a default, replacing the single --config flag). The reproducibility page in particular should explain how a run is now pinned across four files instead of one, and how augmentations are declared declaratively in data_config.yaml and compiled by preprocess_data_config.

The nbs/nb_experiments/Profiler.ipynb markdown also references linear-dinov3-base.yaml / combined_mermaid_coralnet.yaml / run1_train_sources.csv and should be refreshed for consistency, though that's a notebook rather than a wiki page.

Origin: #140. See also #111 (config refactor), #141 (config/registry structural cleanup).