I think the Wikipedia definition is limited, and anybody using the terms keying and line coding interchangably is wrong.
A line code is an encoding (or mapping) from bits to pulse amplitudes (voltages) whose main purposes are spectral shaping and receiver synchronization.
These are some of the things you can acomplish with line coding:
Remove DC. Since long transmission lines do not transfer DC very well, and since DC uses power but transmits no information, many systems use line codes that remove it. The simplest example is AMI (alternate mark inversion). In AMI, a bit 0 is mapped to 0 V, and a bit 1 is mapped to +1 V or -1 V in an alternating fashion.
Spread the signal energy across the entire bandwidth evenly. This is achieved with a scrambler.
Avoid long runs of consecutive '1's or '0's. This helps the receiver synchronize to the received signal; sync algorithmis typically rely on signal transitions. One example is the 8b10b code. This line code substitutes runs of more than 5 symbols by sequences with transitions.
Note that the last two are different ways of achieving similar results.
It is true DC removal is more commonly used in baseband wireline systems. However, line codes, in one shape or another, are used in all communication systems.
Contrast line encoding with other types of codes: source coding compresses the information before transmission, and error control codes add extra information in order to correct and/or detect channel errors. A typical communications system will include all three types of codes.
Regarding keying: keying is the process of taking the output of the line coder and mapping it to a carrier. In this sense, keying does imply a passband channel.