fix(loss): prevent objective-starvation with fixed task weights; target pimt_v3.pt

feat(naturals,shalimar): hierarchical natural oil tokenization

fix(train): keep AdaptiveLossBalancer forward signature intact, expose last_weights attribute

feat(train): log AdaptiveLossBalancer weights per batch for telemetry verification

feat(data): add commercial solvent dilution layer and clean natural profiles

refactor(loss): rename DeepSeekLossBalancer to AdaptiveLossBalancer

feat(train,heads,loss): DeepSeekLossBalancer + continuous cosine alignment

feat(structure,thermo): 3D shape + natural oil reconstruction, 145-D input

refactor(vle): move UNIFAC VLE implementation into pino.thermo module