Some new encoding/decoding options:

• number of bits to write/read per image unit (pixel/DCT coef)
• write/read depth - shift from LSB
• selective filter for JPEG - skip DC coef (first), 0 and 1. Since we have bits and depth params 1 becomes bits << depth
• jpeg-comp - allows to select the JPEG component. By default all components are used
• max-step - overwrite the calculated max step. Also skip size checks in decoder. Can be useful if data in the image is corrupted

All of these options are intended to improve reliability somehow, but if the image is highly optimized, even this is not enough. Especially JPEG, if PNG is not re-encoded, it is lossless.

Also did some refactoring and added tests.