refactor/tensor: refactor synchronization and memory access interface

Remove methods `sync()`, `get()`, `get_mut()`, `remove_copy()` of
`SharedTensor` and introduce new set of methods: `read()`, `read_write()`,
`write_only()`, `drop_device()`. Signature of `SharedTensor::new()` has also
changed.

New API has following benefits:
 - limited checks of use of uninitialized memory,
 - better memory tracking: several memories can be simultaneously marked as
   up-to-date, so some synchronization operations might be skipped,
 - backing memory is automatically allocated on the first use and added to
   `SharedTensor`, even if it's immutable. Mutability is required only for
   reshaping, modifying actual data and dropping memories.

Rationale and design decisions are discussed at the corresponding bugtracker
issue.

BREAKING CHANGE: sync and memory management API of `SharedTensor`

CLOSE: autumnai#37

Cargo.toml

@@ -20,7 +20,6 @@ enum_primitive = "0.1.0"
 byteorder = "0.4"
 num = "0.1"
 lazy_static = "0.1.15"
-linear-map = "0.0.4"
 
 clippy = { version = "0.0.27", optional = true }
 

src/lib.rs

@@ -157,7 +157,6 @@ extern crate enum_primitive;
 extern crate lazy_static;
 extern crate num;
 extern crate byteorder;
-extern crate linear_map;
 
 pub mod backend;
 pub mod device;