Merge task table and task log into a single table (#30)

* Merge task table and task log

* Fix test in db tests

* Address Robert's comments and some better error checking

* Add a LOG_FATAL that exits the program

src/common/doc/tasks.md

@@ -3,6 +3,7 @@
 A *task specification* contains all information that is needed for computing
 the results of a task:
 
+- The ID of the task
 - The function ID of the function that executes the task
 - The arguments (either object IDs for pass by reference
 or values for pass by value)
@@ -11,23 +12,21 @@ or values for pass by value)
 From these, a task ID can be computed which is also stored in the task
 specification.
 
-A *task instance* represents one execution of a task specification.
+A *task* represents the execution of a task specification.
 It consists of:
 
 - A scheduling state (WAITING, SCHEDULED, RUNNING, DONE)
 - The target node where the task is scheduled or executed
-- A unique task instance ID that identifies the particular execution
-  of the task.
+- The task specification
 
 The task data structures are defined in `common/task.h`.
 
-The *task log* is a mapping from the task instance ID to a sequence of
-updates to the status of the task instance. It is updated by various parts
-of the system:
+The *task table* is a mapping from the task ID to the *task* information. It is
+updated by various parts of the system:
 
 1. The local scheduler writes it with status WAITING when submits a task to the global scheduler
 2. The global scheduler appends an update WAITING -> SCHEDULED together with the node ID when assigning the task to a local scheduler
 3. The local scheduler appends an update SCHEDULED -> RUNNING when it assigns a task to a worker
 4. The local scheduler appends an update RUNNING -> DONE when the task finishes execution
 
-The task log is defined in `common/state/task_log.h`.
+The task table is defined in `common/state/task_table.h`.